diff --git a/README.md b/README.md
index db73a1a..5a12835 100644
--- a/README.md
+++ b/README.md
@@ -11,17 +11,18 @@ It currently supports the following local backends:
 - llama-swap
 - oMLX
 - MLX-VLM
+- Any other server with an OpenAI-compatible `/v1/models` endpoint (`openai`, used when nothing more specific matches)
 
-Instead of creating one provider per server, this plugin keeps one `local` provider and lets you register multiple named targets. Each target is probed at runtime, and its currently loaded models are exposed automatically.
+Instead of creating one provider per server, this plugin keeps one `local` provider and lets you register multiple named targets. Each target is probed at runtime, and its models are exposed automatically.
 
 ## Features
 
 - Adds a `local` provider to OpenCode
-- Supports multiple local URLs under one provider
+- Supports multiple servers under one provider
 - Includes supported default `127.0.0.1` targets automatically
-- Detects loaded models at runtime
+- Detects models at runtime and fills in context size, tool calling, and vision support where the server reports them
 - Routes each model to the correct target URL
-- API key auth currently unsupported
+- Optional API key per target, sent as `Authorization: Bearer <key>`
 - Uses OpenCode global config, not project-local config
 
 ## Example
@@ -47,43 +48,49 @@ Default targets are enabled automatically for these backends and ports:
 - Exo: `http://127.0.0.1:52415`
 - llama-swap: `http://127.0.0.1:8080`
 - oMLX: `http://127.0.0.1:8000`
-- MLX-VLM: `http://127.0.0.1:8000`
+- MLX-VLM: `http://127.0.0.1:8080`
 
-If your local providers do not need auth, you can start using the `local` provider immediately.
+If your local servers run on these ports and do not need an API key, you can start using the `local` provider immediately.
 
-**Note: oMLX requires API authentication to be disabled in the web app settings for the plugin to detect and probe models.**
-
-**Note: API key authentication is currently unsupported.**
+**Note: oMLX with API authentication enabled needs a custom target with its API key set (see below).**
 
 ## Custom Targets
 
-If you need non-default hosts or ports, use the CLI auth flow to add an explicit target:
+Add a server on another host or port, or one that needs an API key, with **Add Custom Target**. Run it from the TUI with `/connect` → **Local LLM Provider**, or from the CLI:
 
 ```bash
 opencode auth login --provider local --method "Add Custom Target"
 ```
 
-This will prompt for:
+It asks for:
 
 - a target ID, like `studio` or `remote-ollama`
-- the local provider URL
-- the API key (enter `none` since API keys are currently unsupported)
+- the server URL, with or without `/v1`
+- the server's API key. Leave it empty if the server has none. Enter `{env:VAR_NAME}` to store a reference to an environment variable instead of the key itself
+- which models to list: only loaded models, or every model the server offers (useful for llama-swap, which loads models on demand)
+
+The plugin checks that the server answers before it saves anything, then adds the target to your OpenCode global config. Run it again to add more servers.
 
-The target is then stored in OpenCode global config.
+`/connect` only lists **Local LLM Provider** once the provider has at least one model. If none of your servers are reachable yet, add the first target from the CLI.
 
-You can also add explicit targets manually in config if needed:
+You can also add targets manually in config:
 
 ```json
 {
   "provider": {
     "local": {
-      "name": "Local Provider",
       "options": {
         "includeDefaults": true,
         "targets": {
           "studio": {
             "url": "http://192.168.1.10:1234/v1",
             "kind": "lmstudio"
+          },
+          "swap": {
+            "url": "https://llama-swap.example.com/v1",
+            "kind": "llamaswap",
+            "apiKey": "{env:LLAMA_SWAP_API_KEY}",
+            "includeUnloaded": true
           }
         }
       }
@@ -92,53 +99,34 @@ You can also add explicit targets manually in config if needed:
 }
 ```
 
-Explicit targets override the built-in defaults when they use the same ID.
-The CLI custom-target method is the supported way to add explicit targets without editing config directly.
-
-## Resulting Config
+Target fields:
 
-The plugin stores explicit targets in OpenCode global config under the `local` provider:
-
-```json
-{
-  "provider": {
-    "local": {
-      "name": "Local Provider",
-      "options": {
-        "includeDefaults": true,
-        "targets": {
-          "studio": {
-            "url": "http://127.0.0.1:1234/v1",
-            "kind": "lmstudio"
-          }
-        }
-      }
-    }
-  }
-}
-```
+- `url` (required): the server URL
+- `kind` (optional): one of `ollama`, `lmstudio`, `llamacpp`, `vllm`, `exo`, `llamaswap`, `omlx`, `mlxvlm`, `openai`. If you leave it out, the backend is detected. `openai` treats any server as a plain OpenAI-compatible endpoint
+- `apiKey` (optional): sent as `Authorization: Bearer <key>` when probing the server and with every chat request to its models. Use `{env:VAR_NAME}` to keep the key out of the config file
+- `includeUnloaded` (optional, llama-swap only): list every model llama-swap offers, not only the loaded ones. The plugin reads their details from `/v1/models` and does not load them
 
-With `includeDefaults: true`, the built-in default `127.0.0.1` targets are also checked at runtime even though they are not written into config.
+Explicit targets override the built-in defaults when they use the same ID. A key set in `provider.local.options.apiKey` applies to every target that has no `apiKey` of its own.
 
-**Note: API key authentication is currently unsupported.**
+With `includeDefaults: true`, the built-in default `127.0.0.1` targets are also checked at runtime even though they are not written into config. Set it to `false` if you only use custom targets.
 
 ## How Models Appear
 
-Models are discovered dynamically from each configured target. (Only **loaded** models)
+Models are discovered dynamically from each configured target. By default, only **loaded** models are listed.
 
 To avoid collisions, model IDs are prefixed with the target ID:
 
 - `ollama/llama3.2`
 - `studio/qwen2.5-coder`
 
-Each generated model keeps its own target URL internally, so requests go to the correct backend.
+Each generated model keeps its own target URL and API key internally, so requests go to the correct backend.
 
 ## Notes
 
 - Model detection is runtime-based, not static
 - If loaded models change in your local server, OpenCode will see the updated list on the next provider refresh
 - Built-in default `127.0.0.1` targets are enabled unless you set `includeDefaults` to `false`
-- **API key authentication is currently unsupported**
+- If a custom target cannot be probed, the reason (for example `HTTP 401` for a wrong key) is written to the OpenCode log
 
 ## Development
 
diff --git a/bun.lock b/bun.lock
index e688aa2..3a411d9 100644
--- a/bun.lock
+++ b/bun.lock
@@ -5,8 +5,8 @@
     "": {
       "name": "opencode-local-provider",
       "dependencies": {
-        "@opencode-ai/plugin": "^1.14.46",
-        "@opencode-ai/sdk": "^1.14.46",
+        "@opencode-ai/plugin": "^1.18.31",
+        "@opencode-ai/sdk": "^1.18.31",
         "zod": "^4.4.3",
       },
       "devDependencies": {
@@ -18,21 +18,23 @@
     },
   },
   "packages": {
-    "@msgpackr-extract/msgpackr-extract-darwin-arm64": ["@msgpackr-extract/msgpackr-extract-darwin-arm64@3.0.3", "", { "os": "darwin", "cpu": "arm64" }, "sha512-QZHtlVgbAdy2zAqNA9Gu1UpIuI8Xvsd1v8ic6B2pZmeFnFcMWiPLfWXh7TVw4eGEZ/C9TH281KwhVoeQUKbyjw=="],
+    "@ai-sdk/provider": ["@ai-sdk/provider@3.0.8", "", { "dependencies": { "json-schema": "^0.4.0" } }, "sha512-oGMAgGoQdBXbZqNG0Ze56CHjDZ1IDYOwGYxYjO5KLSlz5HiNQ9udIXsPZ61VWaHGZ5XW/jyjmr6t2xz2jGVwbQ=="],
 
-    "@msgpackr-extract/msgpackr-extract-darwin-x64": ["@msgpackr-extract/msgpackr-extract-darwin-x64@3.0.3", "", { "os": "darwin", "cpu": "x64" }, "sha512-mdzd3AVzYKuUmiWOQ8GNhl64/IoFGol569zNRdkLReh6LRLHOXxU4U8eq0JwaD8iFHdVGqSy4IjFL4reoWCDFw=="],
+    "@msgpackr-extract/msgpackr-extract-darwin-arm64": ["@msgpackr-extract/msgpackr-extract-darwin-arm64@3.0.4", "", { "os": "darwin", "cpu": "arm64" }, "sha512-LCkGo6JDfaBhgST7UpPWgNgLINpcpabaHfyz5OBx75nUYxBsaEPxjnyNjWpeb/xBup/682QnBfRBy2/LvPutZQ=="],
 
-    "@msgpackr-extract/msgpackr-extract-linux-arm": ["@msgpackr-extract/msgpackr-extract-linux-arm@3.0.3", "", { "os": "linux", "cpu": "arm" }, "sha512-fg0uy/dG/nZEXfYilKoRe7yALaNmHoYeIoJuJ7KJ+YyU2bvY8vPv27f7UKhGRpY6euFYqEVhxCFZgAUNQBM3nw=="],
+    "@msgpackr-extract/msgpackr-extract-darwin-x64": ["@msgpackr-extract/msgpackr-extract-darwin-x64@3.0.4", "", { "os": "darwin", "cpu": "x64" }, "sha512-zExlW9zUJKZH/tOtVMttwjKa4Xm/3KcNjnE3dPN92uCktwavMxpgCA3MoJK/DOnTWsQgo224OaST27/mPNAf+w=="],
 
-    "@msgpackr-extract/msgpackr-extract-linux-arm64": ["@msgpackr-extract/msgpackr-extract-linux-arm64@3.0.3", "", { "os": "linux", "cpu": "arm64" }, "sha512-YxQL+ax0XqBJDZiKimS2XQaf+2wDGVa1enVRGzEvLLVFeqa5kx2bWbtcSXgsxjQB7nRqqIGFIcLteF/sHeVtQg=="],
+    "@msgpackr-extract/msgpackr-extract-linux-arm": ["@msgpackr-extract/msgpackr-extract-linux-arm@3.0.4", "", { "os": "linux", "cpu": "arm" }, "sha512-Tg3yX65f5GbtXLkrYEHE5oibZG9epyYWas7FogTTEJeDEF9JlXJzKgXaNhT3UXlTOeA+AfZpYZYZ0uPj7Cfquw=="],
 
-    "@msgpackr-extract/msgpackr-extract-linux-x64": ["@msgpackr-extract/msgpackr-extract-linux-x64@3.0.3", "", { "os": "linux", "cpu": "x64" }, "sha512-cvwNfbP07pKUfq1uH+S6KJ7dT9K8WOE4ZiAcsrSes+UY55E/0jLYc+vq+DO7jlmqRb5zAggExKm0H7O/CBaesg=="],
+    "@msgpackr-extract/msgpackr-extract-linux-arm64": ["@msgpackr-extract/msgpackr-extract-linux-arm64@3.0.4", "", { "os": "linux", "cpu": "arm64" }, "sha512-dgX0P/9wGPJeHFBG+ZmhgE6bmtMt7NP5CRBGyyktpopdk/mW4POnrpQsSLtKI1dwpc+pPLuXHDh6vvskyQE/sw=="],
 
-    "@msgpackr-extract/msgpackr-extract-win32-x64": ["@msgpackr-extract/msgpackr-extract-win32-x64@3.0.3", "", { "os": "win32", "cpu": "x64" }, "sha512-x0fWaQtYp4E6sktbsdAqnehxDgEc/VwM7uLsRCYWaiGu0ykYdZPiS8zCWdnjHwyiumousxfBm4SO31eXqwEZhQ=="],
+    "@msgpackr-extract/msgpackr-extract-linux-x64": ["@msgpackr-extract/msgpackr-extract-linux-x64@3.0.4", "", { "os": "linux", "cpu": "x64" }, "sha512-8TNXMEjJc3QEy7R/x1INhgiU+XakDAFUzBhaz7+Rbrs8NH5UQeHQxxmzsSBJGyV6I1jW79undiQm8tOI+D+8FQ=="],
 
-    "@opencode-ai/plugin": ["@opencode-ai/plugin@1.14.46", "", { "dependencies": { "@opencode-ai/sdk": "1.14.46", "effect": "4.0.0-beta.59", "zod": "4.1.8" }, "peerDependencies": { "@opentui/core": ">=0.2.6", "@opentui/keymap": ">=0.2.6", "@opentui/solid": ">=0.2.6" }, "optionalPeers": ["@opentui/core", "@opentui/keymap", "@opentui/solid"] }, "sha512-JXV0h/dMFGVmEvlKSrbkMeiiIafwWlAGTbAQ4KaK1Bn5yWCFx5DkYVcS3BvgebdJEPkQ4tjVKmG/LHjfZwbjkA=="],
+    "@msgpackr-extract/msgpackr-extract-win32-x64": ["@msgpackr-extract/msgpackr-extract-win32-x64@3.0.4", "", { "os": "win32", "cpu": "x64" }, "sha512-CmCXPQrkbwExx3j946/PtHWHbYJiCRBRDl4BlkRQcJB/YOwQxJRTpoo7aTsortjgoJ1x7opzTSxn7C+ASSLVjQ=="],
 
-    "@opencode-ai/sdk": ["@opencode-ai/sdk@1.14.46", "", { "dependencies": { "cross-spawn": "7.0.6" } }, "sha512-7KOMuoCkNI+bLOw3GCg0nWZ5m7A/MzNsyLfTbZYmE/DIaUqkV2LNRULtrW6PHL1WtYVmJEFPws4dbw/4dVxjzA=="],
+    "@opencode-ai/plugin": ["@opencode-ai/plugin@1.18.31", "", { "dependencies": { "@ai-sdk/provider": "3.0.8", "@opencode-ai/sdk": "1.18.31", "effect": "4.0.0-beta.83", "zod": "4.1.8" }, "peerDependencies": { "@opentui/core": ">=0.4.5", "@opentui/keymap": ">=0.4.5", "@opentui/solid": ">=0.4.5" }, "optionalPeers": ["@opentui/core", "@opentui/keymap", "@opentui/solid"] }, "sha512-Rdc1bPK06PByaGyGd0kf7JUZ4pTkexz2OOUNlqZWLpHOiMEZ+/rFGjt46ypZ3QwA697gNdWwwwQbHbKG5NMwGA=="],
+
+    "@opencode-ai/sdk": ["@opencode-ai/sdk@1.18.31", "", { "dependencies": { "cross-spawn": "7.0.6" } }, "sha512-Raouthf8Lhe9edjvYeeSK7SgvdoU6bBjH9qV3f70dHoa6h+z0X2TMz/e22/wKp/StlFUZ4kIRpYYxFnY8/k01w=="],
 
     "@standard-schema/spec": ["@standard-schema/spec@1.1.0", "", {}, "sha512-l2aFy5jALhniG5HgqrD6jXLi/rUWrKvqN/qJx6yoJsgKhblVd+iqqU4RCXavm/jPityDo5TCvKMnpjKnOriy0w=="],
 
@@ -46,21 +48,23 @@
 
     "detect-libc": ["detect-libc@2.1.2", "", {}, "sha512-Btj2BOOO83o3WyH59e8MgXsxEQVcarkUOpEYrubB0urwnN10yQ364rsiByU11nZlqWYZm05i/of7io4mzihBtQ=="],
 
-    "effect": ["effect@4.0.0-beta.59", "", { "dependencies": { "@standard-schema/spec": "^1.1.0", "fast-check": "^4.6.0", "find-my-way-ts": "^0.1.6", "ini": "^6.0.0", "kubernetes-types": "^1.30.0", "msgpackr": "^1.11.9", "multipasta": "^0.2.7", "toml": "^4.1.1", "uuid": "^13.0.0", "yaml": "^2.8.3" } }, "sha512-xyUDLeHSe8d6lWGOvR6Fgn2HL6gYeTZ/S4Jzk9uc4ZUxMPPsNZlNXrvk0C7/utQFzeX7uAWcVnG2BjbA0SRoAA=="],
+    "effect": ["effect@4.0.0-beta.83", "", { "dependencies": { "@standard-schema/spec": "^1.1.0", "fast-check": "^4.8.0", "find-my-way-ts": "^0.1.6", "ini": "^7.0.0", "kubernetes-types": "^1.30.0", "msgpackr": "^2.0.1", "multipasta": "^0.2.7", "toml": "^4.1.1", "uuid": "^14.0.0", "yaml": "^2.9.0" } }, "sha512-0wsak8RtgGAr9UWSbVDgJHZcUqMSvicHcvaZv1MbMM7MCGgW4Rn/137J1MHQbwYPcwYGxT/IqehFd+UbYuj78w=="],
 
-    "fast-check": ["fast-check@4.7.0", "", { "dependencies": { "pure-rand": "^8.0.0" } }, "sha512-NsZRtqvSSoCP0HbNjUD+r1JH8zqZalyp6gLY9e7OYs7NK9b6AHOs2baBFeBG7bVNsuoukh89x2Yg3rPsul8ziQ=="],
+    "fast-check": ["fast-check@4.10.1", "", { "dependencies": { "pure-rand": "^8.0.0" } }, "sha512-sB5Vghiu8MyCyToHoBVGsT0baZg3sZWNIY+a6Ct2EDrQJlT4YdH6MC1BSLNe3kX1k5i5g0q1O52XFgcKK/rGHg=="],
 
     "find-my-way-ts": ["find-my-way-ts@0.1.6", "", {}, "sha512-a85L9ZoXtNAey3Y6Z+eBWW658kO/MwR7zIafkIUPUMf3isZG0NCs2pjW2wtjxAKuJPxMAsHUIP4ZPGv0o5gyTA=="],
 
-    "ini": ["ini@6.0.0", "", {}, "sha512-IBTdIkzZNOpqm7q3dRqJvMaldXjDHWkEDfrwGEQTs5eaQMWV+djAhR+wahyNNMAa+qpbDUhBMVt4ZKNwpPm7xQ=="],
+    "ini": ["ini@7.0.0", "", {}, "sha512-ifK0CgjALofS5bkrcTy4RaQ9Vx2Knf/eLeIO+NaswQEpH1UblrtTSCIvN71qQDMq0PeQ/SSPojvEJp9vvvfr+w=="],
 
     "isexe": ["isexe@2.0.0", "", {}, "sha512-RHxMLp9lnKHGHRng9QFhRCMbYAcVpn69smSGcq3f36xjgVVWThj4qqLbTLlq7Ssj8B+fIQ1EuCEGI2lKsyQeIw=="],
 
+    "json-schema": ["json-schema@0.4.0", "", {}, "sha512-es94M3nTIfsEPisRafak+HDLfHXnKBhV3vU5eqPcS3flIWqcxJWgXHXiey3YrpaNsanY5ei1VoYEbOzijuq9BA=="],
+
     "kubernetes-types": ["kubernetes-types@1.30.0", "", {}, "sha512-Dew1okvhM/SQcIa2rcgujNndZwU8VnSapDgdxlYoB84ZlpAD43U6KLAFqYo17ykSFGHNPrg0qry0bP+GJd9v7Q=="],
 
-    "msgpackr": ["msgpackr@1.11.10", "", { "optionalDependencies": { "msgpackr-extract": "^3.0.2" } }, "sha512-iCZNq+HszvF+fC3anCm4nBmWEnbeIAfpDs6IStAEKhQ2YSgkjzVG2FF9XJqwwQh5bH3N9OUTUt4QwVN6MLMLtA=="],
+    "msgpackr": ["msgpackr@2.1.0", "", { "optionalDependencies": { "msgpackr-extract": "^3.0.4" } }, "sha512-p/pBCVO63CsvvpkomUnNNag6+n38rULuDA6HHe70o2gtC8ODI52foF/4ko2qQcp6OiErJXTmrZeXmsGGHsIQNQ=="],
 
-    "msgpackr-extract": ["msgpackr-extract@3.0.3", "", { "dependencies": { "node-gyp-build-optional-packages": "5.2.2" }, "optionalDependencies": { "@msgpackr-extract/msgpackr-extract-darwin-arm64": "3.0.3", "@msgpackr-extract/msgpackr-extract-darwin-x64": "3.0.3", "@msgpackr-extract/msgpackr-extract-linux-arm": "3.0.3", "@msgpackr-extract/msgpackr-extract-linux-arm64": "3.0.3", "@msgpackr-extract/msgpackr-extract-linux-x64": "3.0.3", "@msgpackr-extract/msgpackr-extract-win32-x64": "3.0.3" }, "bin": { "download-msgpackr-prebuilds": "bin/download-prebuilds.js" } }, "sha512-P0efT1C9jIdVRefqjzOQ9Xml57zpOXnIuS+csaB4MdZbTdmGDLo8XhzBG1N7aO11gKDDkJvBLULeFTo46wwreA=="],
+    "msgpackr-extract": ["msgpackr-extract@3.0.4", "", { "dependencies": { "node-gyp-build-optional-packages": "5.2.2" }, "optionalDependencies": { "@msgpackr-extract/msgpackr-extract-darwin-arm64": "3.0.4", "@msgpackr-extract/msgpackr-extract-darwin-x64": "3.0.4", "@msgpackr-extract/msgpackr-extract-linux-arm": "3.0.4", "@msgpackr-extract/msgpackr-extract-linux-arm64": "3.0.4", "@msgpackr-extract/msgpackr-extract-linux-x64": "3.0.4", "@msgpackr-extract/msgpackr-extract-win32-x64": "3.0.4" }, "bin": { "download-msgpackr-prebuilds": "bin/download-prebuilds.js" } }, "sha512-4kmO/MdyUIkLIvTPr8VHLil4AtoKIoniWPIEk5+CDy0xnWC84azhSFmuJ7PxZdsYtiP5kEeQsORAVIeMgxT+Hw=="],
 
     "multipasta": ["multipasta@0.2.7", "", {}, "sha512-KPA58d68KgGil15oDqXjkUBEBYc00XvbPj5/X+dyzeo/lWm9Nc25pQRlf1D+gv4OpK7NM0J1odrbu9JNNGvynA=="],
 
@@ -80,11 +84,11 @@
 
     "undici-types": ["undici-types@7.18.2", "", {}, "sha512-AsuCzffGHJybSaRrmr5eHr81mwJU3kjw6M+uprWvCXiNeN9SOGwQ3Jn8jb8m3Z6izVgknn1R0FTCEAP2QrLY/w=="],
 
-    "uuid": ["uuid@13.0.0", "", { "bin": { "uuid": "dist-node/bin/uuid" } }, "sha512-XQegIaBTVUjSHliKqcnFqYypAd4S+WCYt5NIeRs6w/UAry7z8Y9j5ZwRRL4kzq9U3sD6v+85er9FvkEaBpji2w=="],
+    "uuid": ["uuid@14.0.2", "", { "bin": { "uuid": "dist-node/bin/uuid" } }, "sha512-xZe/16rV4aa+HGSOCiY2YeLT1OybRLrrkL/Rqaq7p7GMVXjFh+6wN4oMYgjFmnSnhY8t6Xpdl2l9qmnHYuMHwQ=="],
 
     "which": ["which@2.0.2", "", { "dependencies": { "isexe": "^2.0.0" }, "bin": { "node-which": "./bin/node-which" } }, "sha512-BLI3Tl1TW3Pvl70l3yq3Y64i+awpwXqsGBYWkkqMtnbXgrMD+yj7rhW0kuEDxzJaYXGjEW5ogapKNMEKNMjibA=="],
 
-    "yaml": ["yaml@2.8.3", "", { "bin": { "yaml": "bin.mjs" } }, "sha512-AvbaCLOO2Otw/lW5bmh9d/WEdcDFdQp2Z2ZUH3pX9U2ihyUY0nvLv7J6TrWowklRGPYbB/IuIMfYgxaCPg5Bpg=="],
+    "yaml": ["yaml@2.9.1", "", { "bin": { "yaml": "bin.mjs" } }, "sha512-3NxN8+78OdzbT7C/WjGsyfPAtJaN3FNDsWxv7Y7mcDsT/oOmgW8BpyQQFFBnvZE3j9Y2Sdz1ULFLezL7Eb2yFw=="],
 
     "zod": ["zod@4.4.3", "", {}, "sha512-ytENFjIJFl2UwYglde2jchW2Hwm4GJFLDiSXWdTrJQBIN9Fcyp7n4DhxJEiWNAJMV1/BqWfW/kkg71UDcHJyTQ=="],
 
diff --git a/package-lock.json b/package-lock.json
index e864cf1..83c744b 100644
--- a/package-lock.json
+++ b/package-lock.json
@@ -9,8 +9,8 @@
       "version": "0.1.8",
       "license": "MIT",
       "dependencies": {
-        "@opencode-ai/plugin": "^1.14.46",
-        "@opencode-ai/sdk": "^1.14.46",
+        "@opencode-ai/plugin": "^1.18.31",
+        "@opencode-ai/sdk": "^1.18.31",
         "zod": "^4.4.3"
       },
       "devDependencies": {
@@ -20,10 +20,22 @@
         "typescript": "^5.9.3"
       }
     },
+    "node_modules/@ai-sdk/provider": {
+      "version": "3.0.8",
+      "resolved": "https://registry.npmjs.org/@ai-sdk/provider/-/provider-3.0.8.tgz",
+      "integrity": "sha512-oGMAgGoQdBXbZqNG0Ze56CHjDZ1IDYOwGYxYjO5KLSlz5HiNQ9udIXsPZ61VWaHGZ5XW/jyjmr6t2xz2jGVwbQ==",
+      "license": "Apache-2.0",
+      "dependencies": {
+        "json-schema": "^0.4.0"
+      },
+      "engines": {
+        "node": ">=18"
+      }
+    },
     "node_modules/@msgpackr-extract/msgpackr-extract-darwin-arm64": {
-      "version": "3.0.3",
-      "resolved": "https://registry.npmjs.org/@msgpackr-extract/msgpackr-extract-darwin-arm64/-/msgpackr-extract-darwin-arm64-3.0.3.tgz",
-      "integrity": "sha512-QZHtlVgbAdy2zAqNA9Gu1UpIuI8Xvsd1v8ic6B2pZmeFnFcMWiPLfWXh7TVw4eGEZ/C9TH281KwhVoeQUKbyjw==",
+      "version": "3.0.4",
+      "resolved": "https://registry.npmjs.org/@msgpackr-extract/msgpackr-extract-darwin-arm64/-/msgpackr-extract-darwin-arm64-3.0.4.tgz",
+      "integrity": "sha512-LCkGo6JDfaBhgST7UpPWgNgLINpcpabaHfyz5OBx75nUYxBsaEPxjnyNjWpeb/xBup/682QnBfRBy2/LvPutZQ==",
       "cpu": [
         "arm64"
       ],
@@ -34,9 +46,9 @@
       ]
     },
     "node_modules/@msgpackr-extract/msgpackr-extract-darwin-x64": {
-      "version": "3.0.3",
-      "resolved": "https://registry.npmjs.org/@msgpackr-extract/msgpackr-extract-darwin-x64/-/msgpackr-extract-darwin-x64-3.0.3.tgz",
-      "integrity": "sha512-mdzd3AVzYKuUmiWOQ8GNhl64/IoFGol569zNRdkLReh6LRLHOXxU4U8eq0JwaD8iFHdVGqSy4IjFL4reoWCDFw==",
+      "version": "3.0.4",
+      "resolved": "https://registry.npmjs.org/@msgpackr-extract/msgpackr-extract-darwin-x64/-/msgpackr-extract-darwin-x64-3.0.4.tgz",
+      "integrity": "sha512-zExlW9zUJKZH/tOtVMttwjKa4Xm/3KcNjnE3dPN92uCktwavMxpgCA3MoJK/DOnTWsQgo224OaST27/mPNAf+w==",
       "cpu": [
         "x64"
       ],
@@ -47,9 +59,9 @@
       ]
     },
     "node_modules/@msgpackr-extract/msgpackr-extract-linux-arm": {
-      "version": "3.0.3",
-      "resolved": "https://registry.npmjs.org/@msgpackr-extract/msgpackr-extract-linux-arm/-/msgpackr-extract-linux-arm-3.0.3.tgz",
-      "integrity": "sha512-fg0uy/dG/nZEXfYilKoRe7yALaNmHoYeIoJuJ7KJ+YyU2bvY8vPv27f7UKhGRpY6euFYqEVhxCFZgAUNQBM3nw==",
+      "version": "3.0.4",
+      "resolved": "https://registry.npmjs.org/@msgpackr-extract/msgpackr-extract-linux-arm/-/msgpackr-extract-linux-arm-3.0.4.tgz",
+      "integrity": "sha512-Tg3yX65f5GbtXLkrYEHE5oibZG9epyYWas7FogTTEJeDEF9JlXJzKgXaNhT3UXlTOeA+AfZpYZYZ0uPj7Cfquw==",
       "cpu": [
         "arm"
       ],
@@ -60,9 +72,9 @@
       ]
     },
     "node_modules/@msgpackr-extract/msgpackr-extract-linux-arm64": {
-      "version": "3.0.3",
-      "resolved": "https://registry.npmjs.org/@msgpackr-extract/msgpackr-extract-linux-arm64/-/msgpackr-extract-linux-arm64-3.0.3.tgz",
-      "integrity": "sha512-YxQL+ax0XqBJDZiKimS2XQaf+2wDGVa1enVRGzEvLLVFeqa5kx2bWbtcSXgsxjQB7nRqqIGFIcLteF/sHeVtQg==",
+      "version": "3.0.4",
+      "resolved": "https://registry.npmjs.org/@msgpackr-extract/msgpackr-extract-linux-arm64/-/msgpackr-extract-linux-arm64-3.0.4.tgz",
+      "integrity": "sha512-dgX0P/9wGPJeHFBG+ZmhgE6bmtMt7NP5CRBGyyktpopdk/mW4POnrpQsSLtKI1dwpc+pPLuXHDh6vvskyQE/sw==",
       "cpu": [
         "arm64"
       ],
@@ -73,9 +85,9 @@
       ]
     },
     "node_modules/@msgpackr-extract/msgpackr-extract-linux-x64": {
-      "version": "3.0.3",
-      "resolved": "https://registry.npmjs.org/@msgpackr-extract/msgpackr-extract-linux-x64/-/msgpackr-extract-linux-x64-3.0.3.tgz",
-      "integrity": "sha512-cvwNfbP07pKUfq1uH+S6KJ7dT9K8WOE4ZiAcsrSes+UY55E/0jLYc+vq+DO7jlmqRb5zAggExKm0H7O/CBaesg==",
+      "version": "3.0.4",
+      "resolved": "https://registry.npmjs.org/@msgpackr-extract/msgpackr-extract-linux-x64/-/msgpackr-extract-linux-x64-3.0.4.tgz",
+      "integrity": "sha512-8TNXMEjJc3QEy7R/x1INhgiU+XakDAFUzBhaz7+Rbrs8NH5UQeHQxxmzsSBJGyV6I1jW79undiQm8tOI+D+8FQ==",
       "cpu": [
         "x64"
       ],
@@ -86,9 +98,9 @@
       ]
     },
     "node_modules/@msgpackr-extract/msgpackr-extract-win32-x64": {
-      "version": "3.0.3",
-      "resolved": "https://registry.npmjs.org/@msgpackr-extract/msgpackr-extract-win32-x64/-/msgpackr-extract-win32-x64-3.0.3.tgz",
-      "integrity": "sha512-x0fWaQtYp4E6sktbsdAqnehxDgEc/VwM7uLsRCYWaiGu0ykYdZPiS8zCWdnjHwyiumousxfBm4SO31eXqwEZhQ==",
+      "version": "3.0.4",
+      "resolved": "https://registry.npmjs.org/@msgpackr-extract/msgpackr-extract-win32-x64/-/msgpackr-extract-win32-x64-3.0.4.tgz",
+      "integrity": "sha512-CmCXPQrkbwExx3j946/PtHWHbYJiCRBRDl4BlkRQcJB/YOwQxJRTpoo7aTsortjgoJ1x7opzTSxn7C+ASSLVjQ==",
       "cpu": [
         "x64"
       ],
@@ -99,19 +111,20 @@
       ]
     },
     "node_modules/@opencode-ai/plugin": {
-      "version": "1.14.46",
-      "resolved": "https://registry.npmjs.org/@opencode-ai/plugin/-/plugin-1.14.46.tgz",
-      "integrity": "sha512-JXV0h/dMFGVmEvlKSrbkMeiiIafwWlAGTbAQ4KaK1Bn5yWCFx5DkYVcS3BvgebdJEPkQ4tjVKmG/LHjfZwbjkA==",
+      "version": "1.18.31",
+      "resolved": "https://registry.npmjs.org/@opencode-ai/plugin/-/plugin-1.18.31.tgz",
+      "integrity": "sha512-Rdc1bPK06PByaGyGd0kf7JUZ4pTkexz2OOUNlqZWLpHOiMEZ+/rFGjt46ypZ3QwA697gNdWwwwQbHbKG5NMwGA==",
       "license": "MIT",
       "dependencies": {
-        "@opencode-ai/sdk": "1.14.46",
-        "effect": "4.0.0-beta.59",
+        "@ai-sdk/provider": "3.0.8",
+        "@opencode-ai/sdk": "1.18.31",
+        "effect": "4.0.0-beta.83",
         "zod": "4.1.8"
       },
       "peerDependencies": {
-        "@opentui/core": ">=0.2.6",
-        "@opentui/keymap": ">=0.2.6",
-        "@opentui/solid": ">=0.2.6"
+        "@opentui/core": ">=0.4.5",
+        "@opentui/keymap": ">=0.4.5",
+        "@opentui/solid": ">=0.4.5"
       },
       "peerDependenciesMeta": {
         "@opentui/core": {
@@ -135,9 +148,9 @@
       }
     },
     "node_modules/@opencode-ai/sdk": {
-      "version": "1.14.46",
-      "resolved": "https://registry.npmjs.org/@opencode-ai/sdk/-/sdk-1.14.46.tgz",
-      "integrity": "sha512-7KOMuoCkNI+bLOw3GCg0nWZ5m7A/MzNsyLfTbZYmE/DIaUqkV2LNRULtrW6PHL1WtYVmJEFPws4dbw/4dVxjzA==",
+      "version": "1.18.31",
+      "resolved": "https://registry.npmjs.org/@opencode-ai/sdk/-/sdk-1.18.31.tgz",
+      "integrity": "sha512-Raouthf8Lhe9edjvYeeSK7SgvdoU6bBjH9qV3f70dHoa6h+z0X2TMz/e22/wKp/StlFUZ4kIRpYYxFnY8/k01w==",
       "license": "MIT",
       "dependencies": {
         "cross-spawn": "7.0.6"
@@ -204,27 +217,27 @@
       }
     },
     "node_modules/effect": {
-      "version": "4.0.0-beta.59",
-      "resolved": "https://registry.npmjs.org/effect/-/effect-4.0.0-beta.59.tgz",
-      "integrity": "sha512-xyUDLeHSe8d6lWGOvR6Fgn2HL6gYeTZ/S4Jzk9uc4ZUxMPPsNZlNXrvk0C7/utQFzeX7uAWcVnG2BjbA0SRoAA==",
+      "version": "4.0.0-beta.83",
+      "resolved": "https://registry.npmjs.org/effect/-/effect-4.0.0-beta.83.tgz",
+      "integrity": "sha512-0wsak8RtgGAr9UWSbVDgJHZcUqMSvicHcvaZv1MbMM7MCGgW4Rn/137J1MHQbwYPcwYGxT/IqehFd+UbYuj78w==",
       "license": "MIT",
       "dependencies": {
         "@standard-schema/spec": "^1.1.0",
-        "fast-check": "^4.6.0",
+        "fast-check": "^4.8.0",
         "find-my-way-ts": "^0.1.6",
-        "ini": "^6.0.0",
+        "ini": "^7.0.0",
         "kubernetes-types": "^1.30.0",
-        "msgpackr": "^1.11.9",
+        "msgpackr": "^2.0.1",
         "multipasta": "^0.2.7",
         "toml": "^4.1.1",
-        "uuid": "^13.0.0",
-        "yaml": "^2.8.3"
+        "uuid": "^14.0.0",
+        "yaml": "^2.9.0"
       }
     },
     "node_modules/fast-check": {
-      "version": "4.7.0",
-      "resolved": "https://registry.npmjs.org/fast-check/-/fast-check-4.7.0.tgz",
-      "integrity": "sha512-NsZRtqvSSoCP0HbNjUD+r1JH8zqZalyp6gLY9e7OYs7NK9b6AHOs2baBFeBG7bVNsuoukh89x2Yg3rPsul8ziQ==",
+      "version": "4.10.1",
+      "resolved": "https://registry.npmjs.org/fast-check/-/fast-check-4.10.1.tgz",
+      "integrity": "sha512-sB5Vghiu8MyCyToHoBVGsT0baZg3sZWNIY+a6Ct2EDrQJlT4YdH6MC1BSLNe3kX1k5i5g0q1O52XFgcKK/rGHg==",
       "funding": [
         {
           "type": "individual",
@@ -250,12 +263,12 @@
       "license": "MIT"
     },
     "node_modules/ini": {
-      "version": "6.0.0",
-      "resolved": "https://registry.npmjs.org/ini/-/ini-6.0.0.tgz",
-      "integrity": "sha512-IBTdIkzZNOpqm7q3dRqJvMaldXjDHWkEDfrwGEQTs5eaQMWV+djAhR+wahyNNMAa+qpbDUhBMVt4ZKNwpPm7xQ==",
+      "version": "7.0.0",
+      "resolved": "https://registry.npmjs.org/ini/-/ini-7.0.0.tgz",
+      "integrity": "sha512-ifK0CgjALofS5bkrcTy4RaQ9Vx2Knf/eLeIO+NaswQEpH1UblrtTSCIvN71qQDMq0PeQ/SSPojvEJp9vvvfr+w==",
       "license": "ISC",
       "engines": {
-        "node": "^20.17.0 || >=22.9.0"
+        "node": "^22.22.2 || ^24.15.0 || >=26.0.0"
       }
     },
     "node_modules/isexe": {
@@ -264,6 +277,12 @@
       "integrity": "sha512-RHxMLp9lnKHGHRng9QFhRCMbYAcVpn69smSGcq3f36xjgVVWThj4qqLbTLlq7Ssj8B+fIQ1EuCEGI2lKsyQeIw==",
       "license": "ISC"
     },
+    "node_modules/json-schema": {
+      "version": "0.4.0",
+      "resolved": "https://registry.npmjs.org/json-schema/-/json-schema-0.4.0.tgz",
+      "integrity": "sha512-es94M3nTIfsEPisRafak+HDLfHXnKBhV3vU5eqPcS3flIWqcxJWgXHXiey3YrpaNsanY5ei1VoYEbOzijuq9BA==",
+      "license": "(AFL-2.1 OR BSD-3-Clause)"
+    },
     "node_modules/kubernetes-types": {
       "version": "1.30.0",
       "resolved": "https://registry.npmjs.org/kubernetes-types/-/kubernetes-types-1.30.0.tgz",
@@ -271,18 +290,18 @@
       "license": "Apache-2.0"
     },
     "node_modules/msgpackr": {
-      "version": "1.11.12",
-      "resolved": "https://registry.npmjs.org/msgpackr/-/msgpackr-1.11.12.tgz",
-      "integrity": "sha512-RBdJ1Un7yGlXWajrkxcSa93nvQ0w4zBf60c0yYv7YtBelP8H2FA7XsfBbMHtXKXUMUxH7zV3Zuozh+kUQWhHvg==",
+      "version": "2.1.0",
+      "resolved": "https://registry.npmjs.org/msgpackr/-/msgpackr-2.1.0.tgz",
+      "integrity": "sha512-p/pBCVO63CsvvpkomUnNNag6+n38rULuDA6HHe70o2gtC8ODI52foF/4ko2qQcp6OiErJXTmrZeXmsGGHsIQNQ==",
       "license": "MIT",
       "optionalDependencies": {
-        "msgpackr-extract": "^3.0.2"
+        "msgpackr-extract": "^3.0.4"
       }
     },
     "node_modules/msgpackr-extract": {
-      "version": "3.0.3",
-      "resolved": "https://registry.npmjs.org/msgpackr-extract/-/msgpackr-extract-3.0.3.tgz",
-      "integrity": "sha512-P0efT1C9jIdVRefqjzOQ9Xml57zpOXnIuS+csaB4MdZbTdmGDLo8XhzBG1N7aO11gKDDkJvBLULeFTo46wwreA==",
+      "version": "3.0.4",
+      "resolved": "https://registry.npmjs.org/msgpackr-extract/-/msgpackr-extract-3.0.4.tgz",
+      "integrity": "sha512-4kmO/MdyUIkLIvTPr8VHLil4AtoKIoniWPIEk5+CDy0xnWC84azhSFmuJ7PxZdsYtiP5kEeQsORAVIeMgxT+Hw==",
       "hasInstallScript": true,
       "license": "MIT",
       "optional": true,
@@ -293,18 +312,18 @@
         "download-msgpackr-prebuilds": "bin/download-prebuilds.js"
       },
       "optionalDependencies": {
-        "@msgpackr-extract/msgpackr-extract-darwin-arm64": "3.0.3",
-        "@msgpackr-extract/msgpackr-extract-darwin-x64": "3.0.3",
-        "@msgpackr-extract/msgpackr-extract-linux-arm": "3.0.3",
-        "@msgpackr-extract/msgpackr-extract-linux-arm64": "3.0.3",
-        "@msgpackr-extract/msgpackr-extract-linux-x64": "3.0.3",
-        "@msgpackr-extract/msgpackr-extract-win32-x64": "3.0.3"
+        "@msgpackr-extract/msgpackr-extract-darwin-arm64": "3.0.4",
+        "@msgpackr-extract/msgpackr-extract-darwin-x64": "3.0.4",
+        "@msgpackr-extract/msgpackr-extract-linux-arm": "3.0.4",
+        "@msgpackr-extract/msgpackr-extract-linux-arm64": "3.0.4",
+        "@msgpackr-extract/msgpackr-extract-linux-x64": "3.0.4",
+        "@msgpackr-extract/msgpackr-extract-win32-x64": "3.0.4"
       }
     },
     "node_modules/multipasta": {
-      "version": "0.2.7",
-      "resolved": "https://registry.npmjs.org/multipasta/-/multipasta-0.2.7.tgz",
-      "integrity": "sha512-KPA58d68KgGil15oDqXjkUBEBYc00XvbPj5/X+dyzeo/lWm9Nc25pQRlf1D+gv4OpK7NM0J1odrbu9JNNGvynA==",
+      "version": "0.2.8",
+      "resolved": "https://registry.npmjs.org/multipasta/-/multipasta-0.2.8.tgz",
+      "integrity": "sha512-ZPWuMKyv0cSO29f7hozp+k6+crZbQijV8ipMvxNxRf2SwtYGTX1ZX89Kd20VV4H9Znonx+EQn+iy1wGQsJ+b+Q==",
       "license": "MIT"
     },
     "node_modules/node-gyp-build-optional-packages": {
@@ -332,9 +351,9 @@
       }
     },
     "node_modules/pure-rand": {
-      "version": "8.4.0",
-      "resolved": "https://registry.npmjs.org/pure-rand/-/pure-rand-8.4.0.tgz",
-      "integrity": "sha512-IoM8YF/jY0hiugFo/wOWqfmarlE6J0wc6fDK1PhftMk7MGhVZl88sZimmqBBFomLOCSmcCCpsfj7wXASCpvK9A==",
+      "version": "8.4.2",
+      "resolved": "https://registry.npmjs.org/pure-rand/-/pure-rand-8.4.2.tgz",
+      "integrity": "sha512-vvuOGgcuPJAirlHvuQw1TrOiw7ptaIXXmIbNuiNOY6lNGJJH49PQ1Kj4nd783nPdQhQdicgOjVI2yI/9BD6/Ng==",
       "funding": [
         {
           "type": "individual",
@@ -369,9 +388,9 @@
       }
     },
     "node_modules/toml": {
-      "version": "4.1.1",
-      "resolved": "https://registry.npmjs.org/toml/-/toml-4.1.1.tgz",
-      "integrity": "sha512-EBJnVBr3dTXdA89WVFoAIPUqkBjxPMwRqsfuo1r240tKFHXv3zgca4+NJib/h6TyvGF7vOawz0jGuryJCdNHrw==",
+      "version": "4.3.0",
+      "resolved": "https://registry.npmjs.org/toml/-/toml-4.3.0.tgz",
+      "integrity": "sha512-lVb8X9BsPVuH0M4BKeS91tXAmJvCjQ5UIyAbQFaxkKGyUFK2RPkhwaFSQH8vbpl1d23eu/IBH+dwVMHWaq9A5A==",
       "license": "MIT",
       "engines": {
         "node": ">=20"
@@ -399,9 +418,9 @@
       "license": "MIT"
     },
     "node_modules/uuid": {
-      "version": "13.0.2",
-      "resolved": "https://registry.npmjs.org/uuid/-/uuid-13.0.2.tgz",
-      "integrity": "sha512-vzi9uRZ926x4XV73S/4qQaTwPXM2JBj6/6lI/byHH1jOpCzb0zDbfytgA9LcN/hzb2l7WQSQnxITOVx5un/wGw==",
+      "version": "14.0.2",
+      "resolved": "https://registry.npmjs.org/uuid/-/uuid-14.0.2.tgz",
+      "integrity": "sha512-xZe/16rV4aa+HGSOCiY2YeLT1OybRLrrkL/Rqaq7p7GMVXjFh+6wN4oMYgjFmnSnhY8t6Xpdl2l9qmnHYuMHwQ==",
       "funding": [
         "https://github.com/sponsors/broofa",
         "https://github.com/sponsors/ctavan"
@@ -427,9 +446,9 @@
       }
     },
     "node_modules/yaml": {
-      "version": "2.8.4",
-      "resolved": "https://registry.npmjs.org/yaml/-/yaml-2.8.4.tgz",
-      "integrity": "sha512-ml/JPOj9fOQK8RNnWojA67GbZ0ApXAUlN2UQclwv2eVgTgn7O9gg9o7paZWKMp4g0H3nTLtS9LVzhkpOFIKzog==",
+      "version": "2.9.1",
+      "resolved": "https://registry.npmjs.org/yaml/-/yaml-2.9.1.tgz",
+      "integrity": "sha512-3NxN8+78OdzbT7C/WjGsyfPAtJaN3FNDsWxv7Y7mcDsT/oOmgW8BpyQQFFBnvZE3j9Y2Sdz1ULFLezL7Eb2yFw==",
       "license": "ISC",
       "bin": {
         "yaml": "bin.mjs"
diff --git a/package.json b/package.json
index a9ed283..d6837d2 100644
--- a/package.json
+++ b/package.json
@@ -48,8 +48,8 @@
     "typescript": "^5.9.3"
   },
   "dependencies": {
-    "@opencode-ai/plugin": "^1.14.46",
-    "@opencode-ai/sdk": "^1.14.46",
+    "@opencode-ai/plugin": "^1.18.31",
+    "@opencode-ai/sdk": "^1.18.31",
     "zod": "^4.4.3"
   }
 }
diff --git a/src/config.ts b/src/config.ts
index 2cca4a9..326ef6c 100644
--- a/src/config.ts
+++ b/src/config.ts
@@ -32,20 +32,19 @@ function createV2OpencodeClient(url: URL, input: PluginInput["client"]) {
   })
 }
 
-export function getAuthApiKey(auth?: { type: string; key?: string }) {
-  if (!auth || auth.type !== "api") return ""
-  return auth.key ?? ""
-}
-
 function parseTargetConfig(item: unknown) {
   if (typeof item === "string" && item) return { url: baseURL(item) }
   if (item && typeof item === "object") {
     const url = "url" in item ? item.url : undefined
     const kind = "kind" in item ? item.kind : undefined
+    const apiKey = "apiKey" in item ? item.apiKey : undefined
+    const includeUnloaded = "includeUnloaded" in item ? item.includeUnloaded : undefined
     if (typeof url === "string" && url) {
       return {
         url: baseURL(url),
         ...(typeof kind === "string" && KINDS.includes(kind as (typeof KINDS)[number]) ? { kind } : {}),
+        ...(typeof apiKey === "string" && apiKey ? { apiKey } : {}),
+        ...(includeUnloaded === true ? { includeUnloaded } : {}),
       }
     }
   }
@@ -55,7 +54,7 @@ const defaults = Object.fromEntries(
   Object.entries(supportedProviderDefaultURLs).map(([id, url]) => [
     id,
     {
-      url: baseURL(url),
+      url: baseURL(url!),
       kind: id as LocalTarget["kind"],
     },
   ]),
@@ -103,47 +102,42 @@ export function getProviderTargets(provider?: Pick<Provider, "options">) {
   }
 }
 
-export function getProviderApiKey(provider?: Pick<Provider, "options">, auth?: { type: string; key?: string }) {
+// A provider-wide `options.apiKey` is what OpenCode itself sends to every
+// model of the provider, so probes fall back to it when a target has no key.
+export function getProviderApiKey(provider?: Pick<Provider, "options">) {
   const val = provider?.options?.apiKey
-  if (typeof val === "string" && val) return val
-  return getAuthApiKey(auth)
+  return typeof val === "string" ? val : ""
 }
 
-export async function getCurrentProviderConfig(url: URL, input: PluginInput["client"]) {
-  const cfg = await createV2OpencodeClient(url, input).global.config.get()
-  const provider = cfg.data?.provider?.[LOCAL_PROVIDER_ID]
-  return {
-    targets: getConfiguredTargets(provider as Pick<Provider, "options"> | undefined),
-    key: typeof provider?.options?.apiKey === "string" ? provider.options.apiKey : "",
-  }
+// Resolve `{env:VAR}` the same way OpenCode does when it loads config, for
+// values that have not been through config loading yet (CLI prompt input).
+export function resolveEnv(value: string) {
+  return value.replace(/\{env:([^}]+)\}/g, (_, name: string) => process.env[name] || "")
 }
 
 export async function saveProviderTarget(
   server: URL,
   input: PluginInput["client"],
   id: string,
-  url: string,
-  kind?: LocalTarget["kind"],
-  key?: string,
+  target: LocalTarget,
 ) {
-  const cur = await getCurrentProviderConfig(server, input)
-  const options: Record<string, unknown> = {
-    targets: {
-      ...cur.targets,
-      [id]: {
-        url: baseURL(url),
-        ...(kind ? { kind } : {}),
-      },
-    },
-  }
-
-  if (key !== undefined) options.apiKey = key
-
+  // The global config update deep-merges into the config file, so only the
+  // new target is sent. Re-sending the others would replace any `{env:VAR}`
+  // references in them with the resolved values.
   await createV2OpencodeClient(server, input).global.config.update({
     config: {
       provider: {
         [LOCAL_PROVIDER_ID]: {
-          options,
+          options: {
+            targets: {
+              [id]: {
+                url: baseURL(target.url),
+                ...(target.kind ? { kind: target.kind } : {}),
+                ...(target.apiKey ? { apiKey: target.apiKey } : {}),
+                includeUnloaded: target.includeUnloaded === true,
+              },
+            },
+          },
         },
       },
     },
diff --git a/src/models.ts b/src/models.ts
index f28c914..a182c84 100644
--- a/src/models.ts
+++ b/src/models.ts
@@ -2,7 +2,7 @@ import type { Model } from "@opencode-ai/sdk/v2"
 
 import { DEFAULT_CONTEXT, DEFAULT_OUTPUT, OPENAI_COMPATIBLE_NPM } from "./constants"
 import type { LocalModel } from "./types"
-import { baseURL } from "./url"
+import { authHeaders, baseURL } from "./url"
 
 type Modality = "text" | "audio" | "image" | "video" | "pdf"
 
@@ -33,8 +33,8 @@ function id(target: string, model: string) {
   return `${target}/${model}`
 }
 
-function name(target: string, model: string) {
-  return `${model} (${target})`
+function name(target: string, model: LocalModel) {
+  return `${model.name ?? model.id} (${target})`
 }
 
 function modalities(model: LocalModel): { input: Modality[]; output: Modality[] } {
@@ -57,7 +57,7 @@ function item(providerID: string, target: string, url: string, model: LocalModel
       url: baseURL(url),
       npm: prev?.api.npm ?? OPENAI_COMPATIBLE_NPM,
     },
-    name: prev?.name ?? name(target, model.id),
+    name: prev?.name ?? name(target, model),
     family: prev?.family ?? "",
     capabilities: {
       temperature: prev?.capabilities.temperature ?? true,
@@ -100,11 +100,11 @@ function item(providerID: string, target: string, url: string, model: LocalModel
   }
 }
 
-function configItem(target: string, url: string, model: LocalModel, prev?: ConfigModel) {
+function configItem(target: string, url: string, model: LocalModel, prev?: ConfigModel, apiKey?: string) {
   const attach = model.vision || prev?.attachment || false
   return {
     id: model.id,
-    name: prev?.name ?? name(target, model.id),
+    name: prev?.name ?? name(target, model),
     family: prev?.family || undefined,
     release_date: prev?.release_date || undefined,
     attachment: attach,
@@ -124,7 +124,14 @@ function configItem(target: string, url: string, model: LocalModel, prev?: Confi
       ...prev?.options,
       target,
     },
-    headers: prev?.headers ?? {},
+    // OpenAI-compatible SDKs send model headers on every request, which is how
+    // each target gets its own key under the single `local` provider. The
+    // target key goes last so a header injected by an earlier run can't
+    // outlive a key change.
+    headers: {
+      ...prev?.headers,
+      ...authHeaders(apiKey),
+    },
     provider: {
       npm: prev?.provider?.npm ?? OPENAI_COMPATIBLE_NPM,
       api: baseURL(url),
@@ -141,11 +148,17 @@ export function build(providerID: string, target: string, url: string, list: Loc
   )
 }
 
-export function buildConfig(target: string, url: string, list: LocalModel[], prev: Record<string, ConfigModel>) {
+export function buildConfig(
+  target: string,
+  url: string,
+  list: LocalModel[],
+  prev: Record<string, ConfigModel>,
+  apiKey?: string,
+) {
   return Object.fromEntries(
     list.map((model) => {
       const key = id(target, model.id)
-      return [key, configItem(target, url, model, prev[key])]
+      return [key, configItem(target, url, model, prev[key], apiKey)]
     }),
   )
 }
diff --git a/src/plugin.ts b/src/plugin.ts
index ea0fbaa..a9ca195 100644
--- a/src/plugin.ts
+++ b/src/plugin.ts
@@ -10,28 +10,42 @@ import {
 } from "./constants"
 import {
   getConfiguredTargets,
+  getProviderApiKey,
   getProviderTargets,
+  resolveEnv,
   saveProviderTarget,
 } from "./config"
 import { buildConfig } from "./models"
 import { supportedProviderKinds } from "./providers"
-import { detect, probe } from "./probe"
+import { probe } from "./probe"
+import type { LocalTarget } from "./types"
 import { trimURL } from "./url"
 
 function validID(value: string) {
   return /^[a-z0-9][a-z0-9-_]*$/.test(value)
 }
 
-async function probeModels(provider?: Pick<Provider, "options" | "models">) {
+function errorMessage(e: unknown) {
+  return e instanceof Error ? e.message : String(e)
+}
+
+async function probeModels(
+  provider?: Pick<Provider, "options" | "models">,
+  onError?: (target: string, e: unknown) => Promise<void>,
+) {
   const list = getProviderTargets(provider)
   if (!Object.keys(list).length) return {}
+  const sharedKey = getProviderApiKey(provider)
 
   const all = await Promise.all(
     Object.entries(list).map(async ([id, item]) => {
       try {
-        const found = await probe(item.url, item.kind)
-        return buildConfig(id, item.url, found.models, (provider?.models ?? {}) as Record<string, never>)
-      } catch {
+        const found = await probe(item.url, item.kind, { ...item, apiKey: item.apiKey || sharedKey })
+        // OpenAI-compatible SDKs already send the provider-wide key, so only a
+        // target's own key needs to be added to its models.
+        return buildConfig(id, item.url, found.models, (provider?.models ?? {}) as Record<string, never>, item.apiKey)
+      } catch (e) {
+        await onError?.(id, e)
         return {}
       }
     }),
@@ -41,13 +55,16 @@ async function probeModels(provider?: Pick<Provider, "options" | "models">) {
 }
 
 export const LocalProviderPlugin: Plugin = async (ctx) => {
-  await ctx.client.app.log({
-    body: {
-      service: LOCAL_PLUGIN_SERVICE,
-      level: "info",
-      message: `Local Provider plugin loaded v${pkg.version}. Supported backends: ${supportedProviderKinds.join(", ")}`,
-    },
-  })
+  const log = (level: "info" | "warn" | "error", message: string) =>
+    ctx.client.app.log({
+      body: {
+        service: LOCAL_PLUGIN_SERVICE,
+        level,
+        message,
+      },
+    })
+
+  await log("info", `Local Provider plugin loaded v${pkg.version}. Supported backends: ${supportedProviderKinds.join(", ")}`)
 
   return {
     config: async (cfg) => {
@@ -65,17 +82,29 @@ export const LocalProviderPlugin: Plugin = async (ctx) => {
         name: provider.name ?? LOCAL_PROVIDER_NAME,
         npm: provider.npm ?? OPENAI_COMPATIBLE_NPM,
         options,
-        models: await probeModels({
-          options,
-          models: (provider as Provider).models,
-        }),
+        models: await probeModels(
+          {
+            options,
+            models: (provider as Provider).models,
+          },
+          async (target, e) => {
+            // Built-in default targets are expected to be offline most of the time.
+            if (!list[target]) return
+            await log("warn", `Target "${target}" probe failed: ${errorMessage(e)}`)
+          },
+        ),
       }
     },
     auth: {
       provider: LOCAL_PROVIDER_ID,
       methods: [
         {
-          type: "api",
+          // "oauth" is the only method type OpenCode passes to the plugin from
+          // both /connect and `opencode auth login`. /connect saves an "api"
+          // method's answers without calling the plugin, and the CLI adds an
+          // API key prompt the plugin never sees. "auto" finishes without a
+          // browser step.
+          type: "oauth",
           label: "Add Custom Target",
           prompts: [
             {
@@ -91,40 +120,63 @@ export const LocalProviderPlugin: Plugin = async (ctx) => {
             {
               type: "text",
               key: "baseURL",
-              message: "Enter your local provider URL",
+              message: "Enter the server URL",
               placeholder: "http://192.168.1.10:1234",
               validate(value) {
                 if (!trimURL(value ?? "")) return "URL is required"
               },
             },
+            {
+              type: "text",
+              key: "apiKey",
+              message: "Enter the server's API key, or {env:VAR} to read it from the environment. Leave empty if it has none",
+              placeholder: "no key",
+            },
+            {
+              type: "select",
+              key: "models",
+              message: "Which models should be listed?",
+              options: [
+                { label: "Loaded models only", value: "loaded" },
+                { label: "All models the server offers", value: "all", hint: "llama-swap loads them on demand" },
+              ],
+            },
           ],
           async authorize(input = {}) {
             const id = input.target?.trim() ?? ""
-            const raw = trimURL(input.baseURL ?? "")
-            try {
-              if (!id || !validID(id) || !raw) {
-                throw new Error("Invalid target ID or URL")
-              }
-
-              const result = await probe(raw)
-              const kind = result.kind
-              await saveProviderTarget(ctx.serverUrl, ctx.client, id, raw, kind)
+            const url = trimURL(input.baseURL ?? "")
+            const entered = input.apiKey?.trim() ?? ""
+            const apiKey = entered.toLowerCase() === "none" ? "" : entered
+            const includeUnloaded = input.models === "all"
 
-              return {
-                type: "success" as const,
-                provider: LOCAL_PROVIDER_ID,
-                key: "",
-              }
+            let target: LocalTarget
+            try {
+              if (!validID(id)) throw new Error("Target ID must use lowercase letters, numbers, - or _")
+              if (!url) throw new Error("URL is required")
+              // Probe before saving so a wrong URL or key fails right away.
+              const found = await probe(url, undefined, { apiKey: resolveEnv(apiKey), includeUnloaded })
+              target = { url, kind: found.kind, apiKey, includeUnloaded }
             } catch (e) {
-              const errorMessage = e instanceof Error ? e.message : String(e)
-              await ctx.client.app.log({
-                body: {
-                  service: LOCAL_PLUGIN_SERVICE,
-                  level: "error",
-                  message: `Authorization failed: ${errorMessage}`,
-                },
-              })
-              return { type: "failed" as const }
+              await log("error", `Add Custom Target failed: ${errorMessage(e)}`)
+              throw e
+            }
+
+            return {
+              url: "",
+              method: "auto" as const,
+              instructions: `Found ${target.kind} at ${url}. Saving target "${id}".`,
+              // Saving config restarts OpenCode's instances. Doing it here
+              // rather than in authorize() keeps /connect's pending login
+              // alive until this callback has been looked up.
+              async callback() {
+                try {
+                  await saveProviderTarget(ctx.serverUrl, ctx.client, id, target)
+                  return { type: "success" as const, provider: LOCAL_PROVIDER_ID, key: "" }
+                } catch (e) {
+                  await log("error", `Saving target "${id}" failed: ${errorMessage(e)}`)
+                  return { type: "failed" as const }
+                }
+              },
             }
           },
         },
diff --git a/src/probe.ts b/src/probe.ts
index 793ea8a..95ff064 100644
--- a/src/probe.ts
+++ b/src/probe.ts
@@ -1,23 +1,48 @@
-import type { LocalProviderKind } from "./types"
+import type { LocalProviderKind, LocalTarget } from "./types"
 
 import { supportedProviders, supportedProviderKinds } from "./providers"
-import { rootURL } from "./url"
+import { authHeaders, rootURL } from "./url"
 
-export async function detect(url: string): Promise<LocalProviderKind | null> {
+export type ProbeTarget = Pick<LocalTarget, "apiKey" | "includeUnloaded">
+
+export async function detect(url: string, apiKey?: string): Promise<LocalProviderKind | null> {
   const root = rootURL(url)
+  const headers = authHeaders(apiKey)
 
   for (const kind of supportedProviderKinds) {
-    if (await supportedProviders[kind].detect(root)) return kind
+    if (await supportedProviders[kind].detect(root, headers)) return kind
   }
 
   return null
 }
 
-export async function probe(url: string, kind?: LocalProviderKind) {
+// Every detector swallows errors, so a rejected key looks like an unknown
+// server. Check once more to tell the two apart in the error.
+async function notDetected(url: string, root: string, headers: Record<string, string>) {
+  const status = await fetch(root + "/v1/models", {
+    headers,
+    signal: AbortSignal.timeout(1000),
+  }).then(
+    (res) => res.status,
+    () => undefined,
+  )
+  if (status === 401 || status === 403) {
+    return new Error(`${url} rejected the request (HTTP ${status}), check the target's API key`)
+  }
+  return new Error(`No supported local provider detected at: ${url}`)
+}
+
+export async function probe(url: string, kind?: LocalProviderKind, target: ProbeTarget = {}) {
   const root = rootURL(url)
-  const detected = await detect(root)
+  const headers = authHeaders(target.apiKey)
+  // Every backend also matches the generic "openai" kind, so asking for it
+  // explicitly skips detection of the specific backends.
+  const detected =
+    kind === "openai"
+      ? (await supportedProviders.openai.detect(root, headers)) ? kind : null
+      : await detect(root, target.apiKey)
 
-  if (!detected) throw new Error(`No supported local provider detected at: ${url}`)
+  if (!detected) throw await notDetected(url, root, headers)
 
   if (kind && detected !== kind) {
     throw new Error(`Expected ${kind} at ${url} but detected ${detected}`)
@@ -25,6 +50,6 @@ export async function probe(url: string, kind?: LocalProviderKind) {
 
   return {
     kind: detected,
-    models: await supportedProviders[detected].probe(root),
+    models: await supportedProviders[detected].probe(root, headers, { url, ...target }),
   }
 }
diff --git a/src/providers/exo.ts b/src/providers/exo.ts
index 79cef67..9eb77e0 100644
--- a/src/providers/exo.ts
+++ b/src/providers/exo.ts
@@ -54,9 +54,10 @@ const ModelsResponseSchema = z.object({
     .optional(),
 })
 
-async function detect(url: string) {
+async function detect(url: string, headers?: Record<string, string>) {
   try {
     const res = await fetch(url + "/v1/models", {
+      headers,
       signal: AbortSignal.timeout(1000),
     })
     if (!res.ok) return false
@@ -73,8 +74,9 @@ async function detect(url: string) {
   }
 }
 
-async function probe(url: string): Promise<LocalModel[]> {
+async function probe(url: string, headers?: Record<string, string>): Promise<LocalModel[]> {
   const res = await fetch(url + "/state", {
+    headers,
     signal: AbortSignal.timeout(1000),
   })
   if (!res.ok) throw new Error(`Exo probe failed: ${res.status}`)
diff --git a/src/providers/index.ts b/src/providers/index.ts
index 85afb4b..2cc4334 100644
--- a/src/providers/index.ts
+++ b/src/providers/index.ts
@@ -7,6 +7,7 @@ import lmstudio from "./lmstudio"
 import mlxvlm from "./mlxvlm"
 import ollama from "./ollama"
 import omlx from "./omlx"
+import openai from "./openai"
 import type { ProviderMap } from "./shared"
 import vllm from "./vllm"
 
@@ -19,9 +20,12 @@ export const supportedProviders: ProviderMap = {
   llamaswap,
   omlx,
   mlxvlm,
+  openai,
 }
 
-export const supportedProviderDefaultURLs: Record<LocalProviderKind, string> = {
+// The generic OpenAI-compatible fallback has no well-known port, so it is
+// only used for explicit targets.
+export const supportedProviderDefaultURLs: Partial<Record<LocalProviderKind, string>> = {
   ollama: "http://127.0.0.1:11434",
   lmstudio: "http://127.0.0.1:1234",
   llamacpp: "http://127.0.0.1:8080",
diff --git a/src/providers/llamacpp.ts b/src/providers/llamacpp.ts
index b4d9ab9..710587a 100644
--- a/src/providers/llamacpp.ts
+++ b/src/providers/llamacpp.ts
@@ -21,9 +21,10 @@ const ModelsResponseSchema = z.object({
     .optional(),
 })
 
-export async function runtimeContext(url: string) {
+export async function runtimeContext(url: string, headers?: Record<string, string>) {
   try {
     const propsRes = await fetch(url + "/props", {
+      headers,
       signal: AbortSignal.timeout(1000),
     })
     if (propsRes.ok) {
@@ -36,6 +37,7 @@ export async function runtimeContext(url: string) {
 
   try {
     const slotsRes = await fetch(url + "/slots", {
+      headers,
       signal: AbortSignal.timeout(1000),
     })
     if (slotsRes.ok) {
@@ -50,9 +52,10 @@ export async function runtimeContext(url: string) {
   return null
 }
 
-async function detect(url: string) {
+async function detect(url: string, headers?: Record<string, string>) {
   try {
     const res = await fetch(url, {
+      headers,
       signal: AbortSignal.timeout(1000),
     })
     if (!res.ok) return false
@@ -62,9 +65,10 @@ async function detect(url: string) {
   }
 }
 
-async function probe(url: string): Promise<LocalModel[]> {
-  const loadedContext = await runtimeContext(url)
+async function probe(url: string, headers?: Record<string, string>): Promise<LocalModel[]> {
+  const loadedContext = await runtimeContext(url, headers)
   const res = await fetch(url + "/v1/models", {
+    headers,
     signal: AbortSignal.timeout(1000),
   })
   if (!res.ok) throw new Error(`llama.cpp probe failed: ${res.status}`)
diff --git a/src/providers/llamaswap.ts b/src/providers/llamaswap.ts
index 7bfae76..f20265d 100644
--- a/src/providers/llamaswap.ts
+++ b/src/providers/llamaswap.ts
@@ -1,18 +1,8 @@
 import { z } from "zod"
-import type { LocalModel } from "../types"
+import type { LocalModel, LocalTarget } from "../types"
 import type { ProviderImpl } from "./shared"
 import { runtimeContext } from "./llamacpp"
-
-const ModelsResponseSchema = z.object({
-  data: z
-    .array(
-      z.object({
-        id: z.string(),
-        owned_by: z.string().optional(),
-      }),
-    )
-    .optional(),
-})
+import { listModels, modelFromListing } from "./openai"
 
 const RunningResponseSchema = z.object({
   running: z
@@ -25,9 +15,10 @@ const RunningResponseSchema = z.object({
     .optional(),
 })
 
-export async function runningModels(url: string) {
+export async function runningModels(url: string, headers?: Record<string, string>) {
   try {
     const res = await fetch(url + "/running", {
+      headers,
       signal: AbortSignal.timeout(1000),
     })
     if (!res.ok) return new Set<string>()
@@ -42,43 +33,30 @@ export async function runningModels(url: string) {
   }
 }
 
-async function detect(url: string) {
+async function detect(url: string, headers?: Record<string, string>) {
   try {
     // llama-swap exposes /v1/models for all configured models regardless of load state.
     // We check owned_by to distinguish it from other OpenAI-compatible proxies.
-    const res = await fetch(url + "/v1/models", {
-      signal: AbortSignal.timeout(1000),
-    })
-    if (!res.ok) return false
-    const parsed = ModelsResponseSchema.safeParse(await res.json())
-    if (!parsed.success) return false
-    const data = parsed.data.data
-    if (!data) return false
+    const data = await listModels(url, headers)
     return data.some((item) => item.owned_by === "llama-swap")
   } catch {
     return false
   }
 }
 
-async function probe(url: string): Promise<LocalModel[]> {
-  const res = await fetch(url + "/v1/models", {
-    signal: AbortSignal.timeout(1000),
-  })
-  if (!res.ok) throw new Error(`llama-swap probe failed: ${res.status}`)
-  const body = ModelsResponseSchema.parse(await res.json())
-  if (!body.data) throw new Error("llama-swap probe failed: no data field")
-  const loadedModels = await runningModels(url)
+async function probe(url: string, headers?: Record<string, string>, target?: LocalTarget): Promise<LocalModel[]> {
+  const listed = await listModels(url, headers)
+  const loadedModels = await runningModels(url, headers)
+  const models = target?.includeUnloaded ? listed : listed.filter((item) => loadedModels.has(item.id))
 
   return Promise.all(
-    body.data.filter((item) => loadedModels.has(item.id)).map(async (item) => {
-      const context = (await runtimeContext(`${url}/upstream/${item.id}`)) ?? 0
-
-      return {
-        id: item.id,
-        context,
-        toolcall: false,
-        vision: false,
-      }
+    models.map(async (item) => {
+      const model = modelFromListing(item)
+      // Querying /upstream makes llama-swap load the model, so only running
+      // models get their live context size. The rest keep the listing metadata.
+      if (!loadedModels.has(item.id)) return model
+      const context = await runtimeContext(`${url}/upstream/${item.id}`, headers)
+      return context ? { ...model, context } : model
     }),
   )
 }
diff --git a/src/providers/lmstudio.ts b/src/providers/lmstudio.ts
index 767852c..a1c2e65 100644
--- a/src/providers/lmstudio.ts
+++ b/src/providers/lmstudio.ts
@@ -18,9 +18,10 @@ const ModelsResponseSchema = z.object({
   data: z.array(ModelItemSchema).optional(),
 })
 
-async function detect(url: string) {
+async function detect(url: string, headers?: Record<string, string>) {
   try {
     const res = await fetch(url + "/lmstudio-greeting", {
+      headers,
       signal: AbortSignal.timeout(1000),
     })
     if (!res.ok) return false
@@ -32,8 +33,9 @@ async function detect(url: string) {
   }
 }
 
-async function probe(url: string): Promise<LocalModel[]> {
+async function probe(url: string, headers?: Record<string, string>): Promise<LocalModel[]> {
   const res = await fetch(url + "/api/v0/models", {
+    headers,
     signal: AbortSignal.timeout(1000),
   })
   if (!res.ok) throw new Error(`LM Studio probe failed: ${res.status}`)
diff --git a/src/providers/mlxvlm.ts b/src/providers/mlxvlm.ts
index 6cafcde..7cd761b 100644
--- a/src/providers/mlxvlm.ts
+++ b/src/providers/mlxvlm.ts
@@ -9,8 +9,9 @@ const HealthSchema = z.object({
   loaded_tool_parser: z.string().nullable().optional(),
 })
 
-async function runtimeHealth(url: string, timeout = 1000) {
+async function runtimeHealth(url: string, headers?: Record<string, string>, timeout = 1000) {
   const res = await fetch(url + "/health", {
+    headers,
     signal: AbortSignal.timeout(timeout),
   })
 
@@ -21,9 +22,10 @@ async function runtimeHealth(url: string, timeout = 1000) {
   return HealthSchema.parse(await res.json())
 }
 
-async function detect(url: string) {
+async function detect(url: string, headers?: Record<string, string>) {
   try {
     const res = await fetch(url + "/health", {
+      headers,
       signal: AbortSignal.timeout(1000),
     })
 
@@ -37,8 +39,8 @@ async function detect(url: string) {
   }
 }
 
-async function probe(url: string): Promise<LocalModel[]> {
-  const health = await runtimeHealth(url)
+async function probe(url: string, headers?: Record<string, string>): Promise<LocalModel[]> {
+  const health = await runtimeHealth(url, headers)
   if (!health) {
     throw new Error("MLX-VLM probe failed: health endpoint unavailable")
   }
diff --git a/src/providers/ollama.ts b/src/providers/ollama.ts
index f17c442..178b0ee 100644
--- a/src/providers/ollama.ts
+++ b/src/providers/ollama.ts
@@ -18,9 +18,10 @@ const ShowResponseSchema = z.object({
   capabilities: z.array(z.string()).optional(),
 })
 
-async function detect(url: string) {
+async function detect(url: string, headers?: Record<string, string>) {
   try {
     const res = await fetch(url, {
+      headers,
       signal: AbortSignal.timeout(1000),
     })
     if (!res.ok) return false
@@ -30,11 +31,12 @@ async function detect(url: string) {
   }
 }
 
-async function show(url: string, model: string) {
+async function show(url: string, model: string, headers?: Record<string, string>) {
   try {
     const res = await fetch(url + "/api/show", {
       method: "POST",
       headers: {
+        ...headers,
         "Content-Type": "application/json",
       },
       body: JSON.stringify({ model }),
@@ -47,8 +49,9 @@ async function show(url: string, model: string) {
   }
 }
 
-async function probe(url: string): Promise<LocalModel[]> {
+async function probe(url: string, headers?: Record<string, string>): Promise<LocalModel[]> {
   const res = await fetch(url + "/api/ps", {
+    headers,
     signal: AbortSignal.timeout(1000),
   })
   if (!res.ok) throw new Error(`Ollama probe failed: ${res.status}`)
@@ -57,7 +60,7 @@ async function probe(url: string): Promise<LocalModel[]> {
 
   return Promise.all(
     body.models.map(async (item) => {
-      const extra = await show(url, item.model)
+      const extra = await show(url, item.model, headers)
       return {
         id: item.name,
         context: item.context_length,
diff --git a/src/providers/omlx.ts b/src/providers/omlx.ts
index b0c30bc..c942e5c 100644
--- a/src/providers/omlx.ts
+++ b/src/providers/omlx.ts
@@ -23,9 +23,10 @@ const StatusResponseSchema = z.object({
   ),
 })
 
-async function detect(url: string) {
+async function detect(url: string, headers?: Record<string, string>) {
   try {
     const res = await fetch(url + "/health", {
+      headers,
       signal: AbortSignal.timeout(1000),
     })
     if (!res.ok) return false
@@ -36,8 +37,9 @@ async function detect(url: string) {
   }
 }
 
-async function probe(url: string): Promise<LocalModel[]> {
+async function probe(url: string, headers?: Record<string, string>): Promise<LocalModel[]> {
   const res = await fetch(url + "/v1/models/status", {
+    headers,
     signal: AbortSignal.timeout(1000),
   })
   if (!res.ok) throw new Error(`oMLX probe failed: ${res.status}`)
diff --git a/src/providers/openai.ts b/src/providers/openai.ts
new file mode 100644
index 0000000..20ae7eb
--- /dev/null
+++ b/src/providers/openai.ts
@@ -0,0 +1,100 @@
+import { z } from "zod"
+import type { LocalModel } from "../types"
+import type { ProviderImpl } from "./shared"
+
+// Servers disagree on where (and whether) they put model metadata in
+// /v1/models, so only `id` is required and everything else is read leniently.
+const ModelItemSchema = z.object({
+  id: z.string(),
+  owned_by: z.unknown().optional(),
+  name: z.unknown().optional(),
+  context_length: z.unknown().optional(),
+  context_window: z.unknown().optional(),
+  max_context_length: z.unknown().optional(),
+  max_model_len: z.unknown().optional(),
+  capabilities: z.unknown().optional(),
+  supported_parameters: z.unknown().optional(),
+  architecture: z.unknown().optional(),
+  meta: z.unknown().optional(),
+})
+
+export const ModelsResponseSchema = z.object({
+  data: z.array(ModelItemSchema),
+})
+
+export type ModelItem = z.infer<typeof ModelItemSchema>
+
+function number(value: unknown) {
+  return typeof value === "number" && value > 0 ? value : undefined
+}
+
+function field(value: unknown, key: string) {
+  return value && typeof value === "object" ? (value as Record<string, unknown>)[key] : undefined
+}
+
+function strings(value: unknown) {
+  return Array.isArray(value) ? value.filter((item): item is string => typeof item === "string") : []
+}
+
+// Capabilities come as a list (["tools", "vision"]) or a flag map ({ vision: true }).
+function capabilityList(value: unknown) {
+  if (Array.isArray(value)) return strings(value)
+  if (value && typeof value === "object") {
+    return Object.entries(value)
+      .filter(([, enabled]) => enabled === true)
+      .map(([key]) => key)
+  }
+  return []
+}
+
+export function modelFromListing(item: ModelItem): LocalModel {
+  const capabilities = capabilityList(item.capabilities)
+  const name = typeof item.name === "string" ? item.name.trim() : ""
+  return {
+    id: item.id,
+    ...(name && name !== item.id ? { name } : {}),
+    context:
+      number(item.context_length) ??
+      number(item.context_window) ??
+      number(item.max_context_length) ??
+      number(item.max_model_len) ??
+      number(field(item.meta, "n_ctx")) ??
+      number(field(item.meta, "n_ctx_train")) ??
+      0,
+    toolcall:
+      capabilities.some((item) => ["tools", "tool_use", "function_calling"].includes(item)) ||
+      strings(item.supported_parameters).includes("tools"),
+    vision:
+      capabilities.includes("vision") ||
+      strings(field(item.architecture, "input_modalities")).includes("image"),
+  }
+}
+
+export async function listModels(url: string, headers?: Record<string, string>) {
+  const res = await fetch(url + "/v1/models", {
+    headers,
+    signal: AbortSignal.timeout(1000),
+  })
+  if (!res.ok) throw new Error(`OpenAI-compatible probe failed: ${res.status}`)
+  return ModelsResponseSchema.parse(await res.json()).data
+}
+
+async function detect(url: string, headers?: Record<string, string>) {
+  try {
+    await listModels(url, headers)
+    return true
+  } catch {
+    return false
+  }
+}
+
+async function probe(url: string, headers?: Record<string, string>): Promise<LocalModel[]> {
+  return (await listModels(url, headers)).map(modelFromListing)
+}
+
+const openai: ProviderImpl = {
+  detect,
+  probe,
+}
+
+export default openai
diff --git a/src/providers/shared.ts b/src/providers/shared.ts
index d95551b..82f84a2 100644
--- a/src/providers/shared.ts
+++ b/src/providers/shared.ts
@@ -1,8 +1,8 @@
-import type { LocalModel, LocalProviderKind } from "../types"
+import type { LocalModel, LocalProviderKind, LocalTarget } from "../types"
 
 export type ProviderImpl = {
-  detect(url: string): Promise<boolean>
-  probe(url: string): Promise<LocalModel[]>
+  detect(url: string, headers?: Record<string, string>): Promise<boolean>
+  probe(url: string, headers?: Record<string, string>, target?: LocalTarget): Promise<LocalModel[]>
 }
 
 export type ProviderMap = Record<LocalProviderKind, ProviderImpl>
diff --git a/src/providers/vllm.ts b/src/providers/vllm.ts
index 9e721a9..898bb12 100644
--- a/src/providers/vllm.ts
+++ b/src/providers/vllm.ts
@@ -14,9 +14,10 @@ const ModelsResponseSchema = z.object({
     .optional(),
 })
 
-async function detect(url: string) {
+async function detect(url: string, headers?: Record<string, string>) {
   try {
     const res = await fetch(url + "/v1/models", {
+      headers,
       signal: AbortSignal.timeout(1000),
     })
     if (!res.ok) return false
@@ -31,8 +32,9 @@ async function detect(url: string) {
   }
 }
 
-async function probe(url: string): Promise<LocalModel[]> {
+async function probe(url: string, headers?: Record<string, string>): Promise<LocalModel[]> {
   const res = await fetch(url + "/v1/models", {
+    headers,
     signal: AbortSignal.timeout(1000),
   })
   if (!res.ok) throw new Error(`vLLM probe failed: ${res.status}`)
diff --git a/src/types.ts b/src/types.ts
index 5a0f645..ca30cb9 100644
--- a/src/types.ts
+++ b/src/types.ts
@@ -1,9 +1,12 @@
-export const KINDS = ["ollama", "lmstudio", "llamacpp", "vllm", "exo", "llamaswap", "omlx", "mlxvlm"] as const
+// Detection runs in this order. "openai" matches any server with a /v1/models
+// listing, so it must stay last.
+export const KINDS = ["ollama", "lmstudio", "llamacpp", "vllm", "exo", "llamaswap", "omlx", "mlxvlm", "openai"] as const
 
 export type LocalProviderKind = (typeof KINDS)[number]
 
 export type LocalModel = {
   id: string
+  name?: string
   context: number
   toolcall: boolean
   vision: boolean
@@ -12,4 +15,7 @@ export type LocalModel = {
 export type LocalTarget = {
   url: string
   kind?: LocalProviderKind
+  apiKey?: string
+  // llama-swap: also list models that are not loaded yet
+  includeUnloaded?: boolean
 }
diff --git a/src/url.ts b/src/url.ts
index 423335f..422590e 100644
--- a/src/url.ts
+++ b/src/url.ts
@@ -13,3 +13,7 @@ export function rootURL(url: string) {
   if (!next) return ""
   return next.endsWith("/v1") ? next.slice(0, -3) : next
 }
+
+export function authHeaders(apiKey?: string): Record<string, string> {
+  return apiKey ? { Authorization: `Bearer ${apiKey}` } : {}
+}
diff --git a/tests/providers.test.ts b/tests/providers.test.ts
index 3e47061..32e6c56 100644
--- a/tests/providers.test.ts
+++ b/tests/providers.test.ts
@@ -94,8 +94,11 @@ afterAll(() => {
   compose?.down()
 }, 120_000)
 
+// "openai" is the generic fallback that every suite server also satisfies.
+const backendKinds = supportedProviderKinds.filter((kind) => kind !== "openai")
+
 test("supported providers list stays in sync", () => {
-  expect(supportedProviderKinds).toEqual(suites.map((item) => item.kind))
+  expect(backendKinds).toEqual(suites.map((item) => item.kind))
 })
 
 describe("provider integration", () => {
@@ -115,7 +118,7 @@ describe("provider integration", () => {
     }, 120_000)
 
     test(`${item.kind} reports the expected context length`, async () => {
-      const result = await probe(item.url(), undefined, item.kind)
+      const result = await probe(item.url(), item.kind)
 
       const model = item.modelID ? result.models.find((entry) => entry.id === item.modelID) : result.models[0]
 
@@ -135,8 +138,16 @@ describe("provider integration", () => {
       }
     }, 120_000)
 
+    test(`${item.kind} also probes as a generic OpenAI-compatible server`, async () => {
+      const result = await probe(item.url(), "openai")
+      expect(result.kind).toBe("openai")
+      if (item.modelID) {
+        expect(result.models.some((entry) => entry.id === item.modelID)).toBe(true)
+      }
+    }, 120_000)
+
     test(`${item.kind} probes when kind is supplied`, async () => {
-      const result = await probe(item.url(), undefined, item.kind)
+      const result = await probe(item.url(), item.kind)
       expect(result.kind).toBe(item.kind)
       if (item.modelID) {
         expect(result.models.some((entry) => entry.id === item.modelID)).toBe(true)
@@ -151,7 +162,7 @@ describe("provider integration", () => {
       for (const [otherKind, otherProvider] of Object.entries(supportedProviders)) {
         if (otherKind === item.kind) continue
         const otherDefaultURL = supportedProviderDefaultURLs[otherKind as LocalProviderKind]
-        if (String(new URL(otherDefaultURL).port) !== String(item.port)) continue
+        if (!otherDefaultURL || String(new URL(otherDefaultURL).port) !== String(item.port)) continue
 
         expect(await otherProvider.detect(rootURL(item.url()))).toBe(false)
       }
@@ -160,10 +171,10 @@ describe("provider integration", () => {
 
   test("probe with explicit kind rejects mismatched server", async () => {
     for (const item of activeSuites) {
-      for (const otherKind of supportedProviderKinds) {
+      for (const otherKind of backendKinds) {
         if (otherKind === item.kind) continue
         const otherDefaultURL = supportedProviderDefaultURLs[otherKind]
-        if (String(new URL(otherDefaultURL).port) !== String(item.port)) continue
+        if (!otherDefaultURL || String(new URL(otherDefaultURL).port) !== String(item.port)) continue
 
         await expect(probe(item.url(), otherKind)).rejects.toThrow()
       }
diff --git a/tests/unit/config.test.ts b/tests/unit/config.test.ts
index 23c5271..14cfb07 100644
--- a/tests/unit/config.test.ts
+++ b/tests/unit/config.test.ts
@@ -36,6 +36,35 @@ test("keeps configured targets explicit-only", () => {
   })
 })
 
+test("keeps a non-empty target api key", () => {
+  const targets = getConfiguredTargets({
+    options: {
+      targets: {
+        swap: {
+          url: "https://swap.example.com",
+          kind: "llamaswap",
+          apiKey: "secret",
+        },
+        open: {
+          url: "http://192.168.1.10:1234",
+          apiKey: "",
+        },
+      },
+    },
+  })
+
+  expect(targets).toEqual({
+    swap: {
+      url: "https://swap.example.com/v1",
+      kind: "llamaswap",
+      apiKey: "secret",
+    },
+    open: {
+      url: "http://192.168.1.10:1234/v1",
+    },
+  })
+})
+
 test("configured targets override defaults and suppress duplicate urls", () => {
   const targets = getProviderTargets({
     options: {
diff --git a/tests/unit/llamaswap.test.ts b/tests/unit/llamaswap.test.ts
index 350a4a4..a780223 100644
--- a/tests/unit/llamaswap.test.ts
+++ b/tests/unit/llamaswap.test.ts
@@ -67,4 +67,97 @@ describe("llama-swap provider", () => {
       globalThis.fetch = originalFetch
     }
   })
+
+  test("lists unloaded models from /v1/models without loading them", async () => {
+    const calls: string[] = []
+    const originalFetch = globalThis.fetch
+
+    globalThis.fetch = mock(async (input: string | URL | Request) => {
+      const url = typeof input === "string" ? input : input instanceof URL ? input.toString() : input.url
+      calls.push(url)
+
+      if (url === "http://llamaswap.test/v1/models") {
+        return Response.json({
+          data: [
+            { id: "loaded", owned_by: "llama-swap", context_length: 4096 },
+            {
+              id: "idle",
+              owned_by: "llama-swap",
+              name: "Idle Model",
+              context_length: 32768,
+              capabilities: { vision: true, function_calling: true },
+              supported_parameters: ["tools", "tool_choice"],
+            },
+          ],
+        })
+      }
+
+      if (url === "http://llamaswap.test/running") {
+        return Response.json({ running: [{ model: "loaded", state: "ready" }] })
+      }
+
+      if (url === "http://llamaswap.test/upstream/loaded/props") {
+        return Response.json({ default_generation_settings: { n_ctx: 2048 } })
+      }
+
+      return new Response(null, { status: 404 })
+    }) as unknown as typeof fetch
+
+    try {
+      const result = await probe("http://llamaswap.test", "llamaswap", { includeUnloaded: true })
+
+      expect(result.models).toEqual([
+        { id: "loaded", context: 2048, toolcall: false, vision: false },
+        { id: "idle", name: "Idle Model", context: 32768, toolcall: true, vision: true },
+      ])
+      expect(calls.some((url) => url.includes("/upstream/idle"))).toBe(false)
+    } finally {
+      globalThis.fetch = originalFetch
+    }
+  })
+
+  test("sends the target API key on every probe request", async () => {
+    const calls: string[] = []
+    const originalFetch = globalThis.fetch
+
+    globalThis.fetch = mock(async (input: string | URL | Request, init?: RequestInit) => {
+      const url = typeof input === "string" ? input : input instanceof URL ? input.toString() : input.url
+      calls.push(url)
+
+      if (new Headers(init?.headers).get("Authorization") !== "Bearer secret") {
+        return new Response(JSON.stringify({ error: "unauthorized" }), { status: 401 })
+      }
+
+      if (url === "http://llamaswap.test/v1/models") {
+        return Response.json({ data: [{ id: "loaded", owned_by: "llama-swap" }] })
+      }
+
+      if (url === "http://llamaswap.test/running") {
+        return Response.json({ running: [{ model: "loaded", state: "ready" }] })
+      }
+
+      if (url === "http://llamaswap.test/upstream/loaded/props") {
+        return Response.json({ default_generation_settings: { n_ctx: 8192 } })
+      }
+
+      return new Response(null, { status: 404 })
+    }) as unknown as typeof fetch
+
+    try {
+      await expect(probe("http://llamaswap.test")).rejects.toThrow(
+        "http://llamaswap.test rejected the request (HTTP 401), check the target's API key",
+      )
+
+      const result = await probe("http://llamaswap.test", "llamaswap", { apiKey: "secret" })
+
+      expect(result.kind).toBe("llamaswap")
+      expect(result.models).toEqual([
+        { id: "loaded", context: 8192, toolcall: false, vision: false },
+      ])
+      expect(calls).toContain("http://llamaswap.test/running")
+      expect(calls).toContain("http://llamaswap.test/upstream/loaded/props")
+    } finally {
+      globalThis.fetch = originalFetch
+    }
+  })
 })
diff --git a/tests/unit/openai.test.ts b/tests/unit/openai.test.ts
new file mode 100644
index 0000000..3872965
--- /dev/null
+++ b/tests/unit/openai.test.ts
@@ -0,0 +1,102 @@
+import { describe, expect, mock, test } from "bun:test"
+
+import { probe } from "../../src/probe"
+import { modelFromListing } from "../../src/providers/openai"
+
+function mockFetch(routes: Record<string, unknown>, calls: string[] = []) {
+  return mock(async (input: string | URL | Request) => {
+    const url = typeof input === "string" ? input : input instanceof URL ? input.toString() : input.url
+    calls.push(url)
+    if (url in routes) return Response.json(routes[url])
+    return new Response(null, { status: 404 })
+  }) as unknown as typeof fetch
+}
+
+describe("OpenAI-compatible model listing", () => {
+  test("reads llama-swap style metadata", () => {
+    expect(
+      modelFromListing({
+        id: "qwen",
+        name: "Qwen 3.6",
+        context_length: 65536,
+        capabilities: { vision: true, function_calling: true },
+        supported_parameters: ["tools", "tool_choice"],
+        architecture: { input_modalities: ["text", "image"] },
+      }),
+    ).toEqual({ id: "qwen", name: "Qwen 3.6", context: 65536, toolcall: true, vision: true })
+  })
+
+  test("reads list capabilities and nested context fields", () => {
+    expect(
+      modelFromListing({
+        id: "gemma",
+        capabilities: ["tools"],
+        meta: { n_ctx_train: 8192 },
+      }),
+    ).toEqual({ id: "gemma", context: 8192, toolcall: true, vision: false })
+    expect(modelFromListing({ id: "vllm-model", max_model_len: 4096 }).context).toBe(4096)
+  })
+
+  test("ignores null or malformed metadata", () => {
+    expect(
+      modelFromListing({
+        id: "plain",
+        name: null,
+        context_length: null,
+        capabilities: "tools",
+        meta: null,
+      }),
+    ).toEqual({ id: "plain", context: 0, toolcall: false, vision: false })
+  })
+})
+
+describe("OpenAI-compatible fallback", () => {
+  test("detects servers that only expose /v1/models", async () => {
+    const originalFetch = globalThis.fetch
+    globalThis.fetch = mockFetch({
+      "http://proxy.test/v1/models": {
+        object: "list",
+        data: [{ id: "gpt-oss", object: "model", owned_by: "someone", context_window: 131072 }],
+      },
+    })
+
+    try {
+      const result = await probe("http://proxy.test/v1")
+      expect(result.kind).toBe("openai")
+      expect(result.models).toEqual([{ id: "gpt-oss", context: 131072, toolcall: false, vision: false }])
+    } finally {
+      globalThis.fetch = originalFetch
+    }
+  })
+
+  test("explicit openai kind skips backend detection", async () => {
+    const calls: string[] = []
+    const originalFetch = globalThis.fetch
+    globalThis.fetch = mockFetch(
+      {
+        "http://llamaswap.test/v1/models": { data: [{ id: "idle", owned_by: "llama-swap" }] },
+      },
+      calls,
+    )
+
+    try {
+      const result = await probe("http://llamaswap.test", "openai")
+      expect(result.kind).toBe("openai")
+      expect(result.models.map((item) => item.id)).toEqual(["idle"])
+      expect(calls.every((url) => url === "http://llamaswap.test/v1/models")).toBe(true)
+    } finally {
+      globalThis.fetch = originalFetch
+    }
+  })
+
+  test("still rejects servers without a model listing", async () => {
+    const originalFetch = globalThis.fetch
+    globalThis.fetch = mockFetch({})
+
+    try {
+      await expect(probe("http://nothing.test")).rejects.toThrow("No supported local provider detected")
+    } finally {
+      globalThis.fetch = originalFetch
+    }
+  })
+})
diff --git a/tests/unit/plugin.test.ts b/tests/unit/plugin.test.ts
new file mode 100644
index 0000000..067eccc
--- /dev/null
+++ b/tests/unit/plugin.test.ts
@@ -0,0 +1,278 @@
+import { describe, expect, mock, test } from "bun:test"
+import type { Config, PluginInput } from "@opencode-ai/plugin"
+
+import { saveProviderTarget } from "../../src/config"
+import { LocalProviderPlugin } from "../../src/plugin"
+
+function fakeClient(fetchImpl?: typeof fetch) {
+  const logs: { level: string; message: string }[] = []
+  const client = {
+    app: {
+      log: async ({ body }: { body: { level: string; message: string } }) => {
+        logs.push(body)
+      },
+    },
+    _client: { getConfig: () => ({ fetch: fetchImpl }) },
+  }
+  return { client: client as unknown as PluginInput["client"], logs }
+}
+
+function swapServer(key: string) {
+  return mock(async (input: string | URL | Request, init?: RequestInit) => {
+    const url = typeof input === "string" ? input : input instanceof URL ? input.toString() : input.url
+    if (new Headers(init?.headers).get("Authorization") !== `Bearer ${key}`) {
+      return new Response(null, { status: 401 })
+    }
+    if (url === "https://swap.example.com/v1/models") {
+      return Response.json({
+        data: [{ id: "qwen", owned_by: "llama-swap", name: "Qwen", context_length: 65536 }],
+      })
+    }
+    if (url === "https://swap.example.com/running") return Response.json({ running: [] })
+    return new Response(null, { status: 404 })
+  }) as unknown as typeof fetch
+}
+
+describe("config hook", () => {
+  test("probes keyed targets and sends the key with their models", async () => {
+    const originalFetch = globalThis.fetch
+    globalThis.fetch = swapServer("secret")
+    const { client } = fakeClient()
+
+    try {
+      const hooks = await LocalProviderPlugin({ client, serverUrl: new URL("http://opencode.test") } as PluginInput)
+      const cfg: Config = {
+        provider: {
+          local: {
+            options: {
+              includeDefaults: false,
+              targets: {
+                swap: { url: "https://swap.example.com", apiKey: "secret", includeUnloaded: true },
+              },
+            },
+          },
+        },
+      }
+
+      await hooks.config!(cfg)
+
+      const model = cfg.provider!.local!.models!["swap/qwen"]!
+      expect(model.name).toBe("Qwen (swap)")
+      expect(model.limit).toEqual({ context: 65536, output: 8192 })
+      expect(model.headers).toEqual({ Authorization: "Bearer secret" })
+      // `provider.api` is read by OpenCode at runtime but missing from the SDK config type.
+      expect(model.provider as unknown).toEqual({ npm: "@ai-sdk/openai-compatible", api: "https://swap.example.com/v1" })
+    } finally {
+      globalThis.fetch = originalFetch
+    }
+  })
+
+  test("falls back to the provider-wide apiKey for probing", async () => {
+    const originalFetch = globalThis.fetch
+    globalThis.fetch = swapServer("shared")
+    const { client } = fakeClient()
+
+    try {
+      const hooks = await LocalProviderPlugin({ client, serverUrl: new URL("http://opencode.test") } as PluginInput)
+      const cfg: Config = {
+        provider: {
+          local: {
+            options: {
+              apiKey: "shared",
+              includeDefaults: false,
+              targets: { swap: { url: "https://swap.example.com", includeUnloaded: true } },
+            },
+          },
+        },
+      }
+
+      await hooks.config!(cfg)
+
+      // OpenAI-compatible SDKs send options.apiKey themselves.
+      expect(cfg.provider!.local!.models!["swap/qwen"]!.headers).toEqual({})
+    } finally {
+      globalThis.fetch = originalFetch
+    }
+  })
+
+  test("logs probe failures for configured targets only", async () => {
+    const originalFetch = globalThis.fetch
+    globalThis.fetch = swapServer("secret")
+    const { client, logs } = fakeClient()
+
+    try {
+      const hooks = await LocalProviderPlugin({ client, serverUrl: new URL("http://opencode.test") } as PluginInput)
+      const cfg: Config = {
+        provider: {
+          local: {
+            options: { targets: { swap: { url: "https://swap.example.com", apiKey: "wrong" } } },
+          },
+        },
+      }
+
+      await hooks.config!(cfg)
+
+      expect(cfg.provider!.local!.models).toEqual({})
+      expect(logs.filter((item) => item.level === "warn").map((item) => item.message)).toEqual([
+        'Target "swap" probe failed: https://swap.example.com/v1 rejected the request (HTTP 401), check the target\'s API key',
+      ])
+    } finally {
+      globalThis.fetch = originalFetch
+    }
+  })
+})
+
+describe("saveProviderTarget", () => {
+  test("patches only the new target into global config", async () => {
+    const bodies: unknown[] = []
+    const fetchImpl = mock(async (input: string | URL | Request) => {
+      const request = input as Request
+      bodies.push(await request.json())
+      return Response.json({})
+    }) as unknown as typeof fetch
+    const { client } = fakeClient(fetchImpl)
+
+    await saveProviderTarget(new URL("http://opencode.test"), client, "swap", {
+      url: "https://swap.example.com",
+      kind: "llamaswap",
+      apiKey: "{env:SWAP_KEY}",
+    })
+
+    expect(bodies).toEqual([
+      {
+        provider: {
+          local: {
+            options: {
+              targets: {
+                swap: {
+                  url: "https://swap.example.com/v1",
+                  kind: "llamaswap",
+                  apiKey: "{env:SWAP_KEY}",
+                  includeUnloaded: false,
+                },
+              },
+            },
+          },
+        },
+      },
+    ])
+  })
+})
+
+describe("Add Custom Target", () => {
+  function configRecorder() {
+    const bodies: any[] = []
+    const fetchImpl = mock(async (input: string | URL | Request) => {
+      bodies.push(await (input as Request).json())
+      return Response.json({})
+    }) as unknown as typeof fetch
+    return { bodies, fetchImpl }
+  }
+
+  async function addTarget(fetchImpl: typeof fetch, inputs: Record<string, string>) {
+    const { client, logs } = fakeClient(fetchImpl)
+    const hooks = await LocalProviderPlugin({ client, serverUrl: new URL("http://opencode.test") } as PluginInput)
+    const method = hooks.auth!.methods[0]!
+    if (method.type !== "oauth") throw new Error("expected an oauth method")
+    return { method, logs, run: () => method.authorize(inputs) }
+  }
+
+  test("saves a keyed target only once the login completes", async () => {
+    const originalFetch = globalThis.fetch
+    globalThis.fetch = swapServer("secret")
+    process.env.TEST_SWAP_KEY = "secret"
+    const { bodies, fetchImpl } = configRecorder()
+
+    try {
+      const { run } = await addTarget(fetchImpl, {
+        target: "swap",
+        baseURL: "https://swap.example.com/",
+        apiKey: "{env:TEST_SWAP_KEY}",
+        models: "all",
+      })
+      const authorization = await run()
+
+      expect(authorization.method).toBe("auto")
+      expect(bodies).toEqual([])
+      if (authorization.method !== "auto") return
+
+      expect(await authorization.callback()).toEqual({ type: "success", provider: "local", key: "" })
+      expect(bodies).toEqual([
+        {
+          provider: {
+            local: {
+              options: {
+                targets: {
+                  swap: {
+                    url: "https://swap.example.com/v1",
+                    kind: "llamaswap",
+                    apiKey: "{env:TEST_SWAP_KEY}",
+                    includeUnloaded: true,
+                  },
+                },
+              },
+            },
+          },
+        },
+      ])
+    } finally {
+      globalThis.fetch = originalFetch
+      delete process.env.TEST_SWAP_KEY
+    }
+  })
+
+  test("saves keyless targets without an apiKey", async () => {
+    const originalFetch = globalThis.fetch
+    globalThis.fetch = mock(async (input: string | URL | Request) => {
+      const url = typeof input === "string" ? input : input instanceof URL ? input.toString() : input.url
+      if (url === "http://studio.test/v1/models") return Response.json({ data: [{ id: "qwen" }] })
+      return new Response(null, { status: 404 })
+    }) as unknown as typeof fetch
+    const { bodies, fetchImpl } = configRecorder()
+
+    try {
+      const { run } = await addTarget(fetchImpl, {
+        target: "studio",
+        baseURL: "http://studio.test",
+        apiKey: "",
+        models: "loaded",
+      })
+      const authorization = await run()
+      if (authorization.method !== "auto") throw new Error("expected auto")
+      await authorization.callback()
+
+      expect(bodies[0].provider.local.options.targets).toEqual({
+        studio: { url: "http://studio.test/v1", kind: "openai", includeUnloaded: false },
+      })
+    } finally {
+      globalThis.fetch = originalFetch
+    }
+  })
+
+  test("fails before saving when the server rejects the key", async () => {
+    const originalFetch = globalThis.fetch
+    globalThis.fetch = swapServer("secret")
+    const { bodies, fetchImpl } = configRecorder()
+
+    try {
+      const { run, logs } = await addTarget(fetchImpl, {
+        target: "swap",
+        baseURL: "https://swap.example.com",
+        apiKey: "wrong",
+        models: "loaded",
+      })
+
+      await expect(run()).rejects.toThrow("rejected the request (HTTP 401)")
+      expect(bodies).toEqual([])
+      expect(logs.some((item) => item.level === "error" && item.message.includes("HTTP 401"))).toBe(true)
+    } finally {
+      globalThis.fetch = originalFetch
+    }
+  })
+
+  test("rejects invalid target IDs", async () => {
+    const { fetchImpl } = configRecorder()
+    const { run } = await addTarget(fetchImpl, { target: "My Server", baseURL: "http://studio.test" })
+    await expect(run()).rejects.toThrow("Target ID must use lowercase letters")
+  })
+})