docs(memory): agents check memory before every answer; one AGENTS.md line for the rest

changelog.mdx

@@ -15,6 +15,10 @@ to get notified on every SDK bump.
 - **Scripted runs refuse those actions unless you allow them.** `predev run` and `predev -p` take `--allow writes,secrets,prod,kill,browser,integrations` (or `all`, or `PREDEV_ALLOW`); they also leave your Chrome and connected accounts alone unless allowed. See [What a run may do beyond this folder](/cli/overview#what-a-run-may-do-beyond-this-folder).
 - pre.dev never changes your agent, shell or git settings or credentials, and command output that looks like a secret is redacted before the agent sees it.
 
+### Memory
+
+- **Other agents check your memory before every answer.** The [pre.dev MCP server](/coding-agent/memory#use-it-in-claude-code-cursor-and-other-agents) now tells connected agents to read your memory before they answer or write anything, code, commit messages and tests included, not only at the start of a task. For agents that don't act on a server's instructions, such as Codex, add the one line on that page to `AGENTS.md`.
+
 ## 2026-10-21
 
 ### Memory

coding-agent/memory.mdx

@@ -133,7 +133,15 @@ The pre.dev MCP server gives other coding agents your memory, on your own accoun
 | [`memory_remember`](/mcp/tools#memory_remember) | Saves one lasting preference, for all your projects or for one project |
 | [`memory_forget`](/mcp/tools#memory_forget) | Removes one entry, archived so you can undo it in pre.dev |
 
-The server tells connected agents to read your memory at the start of a task and to save lasting preferences you state. The memory tools work on every plan, Free included, and use no credits.
+The server tells connected agents to check your memory before they answer or write anything, and to save lasting preferences you state. The memory tools work on every plan, Free included, and use no credits.
+
+Some agents, Codex among them, don't act on a server's instructions. Add this line to the agent's `AGENTS.md` (`CLAUDE.md` for Claude Code) so it checks your memory every time:
+
+```text
+Before you start a task, check pre.dev memory: call memory_list (with project set to the pre.dev project named like this folder) and follow what it returns. When I state a lasting preference, save it with memory_remember.
+```
+
+Name the folder like your pre.dev project (the CLI names a project after its folder), so the agent finds that project's memory.
 
 Add pre.dev to Claude Code:
 

mcp/tools.mdx

@@ -219,7 +219,7 @@ The structured result contains `total` and `batches`. Each batch has `id`, `stat
 
 Without `project`, the result is your memory for all your projects, then the projects that have their own memory, with how many entries each holds. With `project`, it is exactly what pre.dev uses in a chat there: your team's entries (Required ones marked `[REQUIRED]`), the project's memory, then yours. Required team entries win, and project or team memory wins over personal memory when they disagree.
 
-Each entry is one text line that starts with its 24-character id, for example `- [65f0c3a1b2c4d5e6f7a8b9c0] Use pnpm, not npm`. A long list ends with how many entries were left out; find those with `memory_search`. The server's instructions ask agents to call `memory_list` at the start of a task and follow it.
+Each entry is one text line that starts with its 24-character id, for example `- [65f0c3a1b2c4d5e6f7a8b9c0] Use pnpm, not npm`. A long list ends with how many entries were left out; find those with `memory_search`. The server's instructions ask agents to call `memory_list` before they answer or write anything, with the project named like the folder they are in, and to follow it.
 
 ## memory_search
 

skill.md

@@ -26,7 +26,7 @@ pre.dev is the Long Horizon Coding Agent: it plans, builds, verifies and ships s
 | Call a model (AI Gateway): chat, embeddings, images, video, speech | `https://api.pre.dev/v1` with an OpenAI or Anthropic SDK |
 | Plan software before building it | `POST https://api.pre.dev/fast-spec` or `/deep-spec`, or the `fast_spec` MCP tool |
 | Give an agent pre.dev tools, plus the user's own Chrome, profiles, and logins | The pre.dev MCP server: `npx -y @predotdev/mcp setup` signs in to pre.dev and adds it to every coding agent on the Mac (run it with a 10 minute timeout and tell the user what to click). Without Node.js, the hosted server `https://api.pre.dev/mcp` has every tool except the Chrome ones. See https://docs.pre.dev/browser-agents/local.md |
-| Follow what the user told pre.dev to remember, or save a lasting preference they state | The MCP tools `memory_list` (call it at the start of a task), `memory_search`, `memory_remember`, and `memory_forget`. See https://docs.pre.dev/coding-agent/memory.md |
+| Follow what the user told pre.dev to remember, or save a lasting preference they state | The MCP tools `memory_list` (call it before you answer or write anything), `memory_search`, `memory_remember`, and `memory_forget`. See https://docs.pre.dev/coding-agent/memory.md |
 | Take payments in an app built on pre.dev | The Stripe SDK, pointed at `STRIPE_API_HOST` |
 | Build or change code | The coding agent, on the web or in the CLI; `predev run` with `PREDEV_API_KEY` runs one task unattended |
 | Search these docs | Docs MCP server `https://docs.pre.dev/mcp` (no key) |