Shared API credentials + service playbooks for every AI coding agent on your machine. Local-first, no SaaS, no per-project re-paste.
One
~/.trove/directory. Every AI coding agent — Claude Code, Codex, Cursor — in every project on your machine reuses the same API credentials and service playbooks.
No more re-pasting Stripe keys into a fresh .env for every project. No more re-explaining "this service has a quirk that bit me last time" to a fresh agent session. Add a service once at ~/.trove/<service>/; reference it from any project's CLAUDE.md / AGENTS.md / .cursorrules with one line:
@/Users/you/.trove/stripe/module.mdThe AI loads the module on session start. It now knows your Stripe key, the gotchas (amounts in cents, restricted keys over secret keys, idempotency-key requirement), and — if a Stripe MCP server is wired up — calls it directly.
No inject step. No init step. No per-agent shims. Trove is a plain directory; the AI is the runtime.
Per service, one folder holds three layers:
- The playbook — gotchas, billing pitfalls, error-code tables, real code snippets, written from production dogfood. Stripe amounts in cents. Supabase RLS opt-in default. GA4 service-account-as-user-not-IAM. The stuff that gets a senior engineer paged at 2am
- API credentials — keys / tokens / files (for service-account JSON / SSH keys / certs), gitignored at file-mode 600
- MCP server pointers (optional) —
mcp:block per module, so the AI configures the agent's MCP settings on first use
Same module works in every agent that supports absolute-path @-reference. Switch projects: same trove. Switch agents (Claude Code → Codex → Cursor): same trove.
| Trove is NOT replacing | Because |
|---|---|
.env + direnv |
Env files re-fragment per project — every new repo means re-pasting the same Stripe key. They also carry no skill knowledge and no MCP config. Trove is the cross-project home env files never tried to be |
~/.claude/skills/ (and equivalents) |
Agent-specific — locked to Claude Code. Trove modules are referenced by absolute path from any agent that supports @-reference, so the same module serves Claude Code / Codex / Cursor without re-authoring |
1Password CLI / op run |
Credentials only — no playbook, no MCP wiring. Use 1Password as the source-of-truth vault and pull values into trove if you want; the two compose |
| Writing your own skill files per project | Every project re-pays the authoring cost; trove de-duplicates that across N projects, and last_verified makes "is this knowledge still correct?" auditable |
| A SaaS / cloud sync | Local-first by design. Sync via git remote / iCloud Drive / rsync — whichever you already use |
A module is a directory under ~/.trove/:
~/.trove/minimax/
├── module.md # YAML frontmatter (schema, applies_to, optional MCP) + skill body
├── credentials.json # secret values — gitignored, mode 600
└── files/ # (optional) file-type creds: SA JSON / SSH key / cert
On session start, the AI:
- Auto-loads the referenced modules (any agent supporting
@-referenceto an absolute path) - Reads
credentials.jsonandfiles/directly when calling the API — no env export, no shell-history exposure - Consults the playbook body when it hits a code path that has known gotchas
- Updates the project's
CLAUDE.md/AGENTS.md/.cursorrulesif you ask it to add a new module
For services with MCP servers, see the next section.
Most modern services ship MCP servers — Stripe, Supabase, Google Analytics, npm, etc. Each agent has its own MCP config file (~/.claude.json, ~/.cursor/mcp.json, ~/.codex/...). Adding one service to three agents currently means editing three files.
Trove modules can carry an mcp: block declaring the canonical install:
mcp:
type: http
url: https://mcp.stripe.comOr for stdio servers:
mcp:
type: stdio
command: pipx
args: ["run", "google-analytics-mcp"]
env:
GOOGLE_APPLICATION_CREDENTIALS: ${credential.GOOGLE_SERVICE_ACCOUNT_JSON}The ${credential.X} substitution resolves to the field's value (string fields) or absolute path (file fields, SPEC §2.3). The AI merges the block into the agent's MCP config on first use — one place to declare, one place to update, applies to every agent.
v0.2.5 — 24 modules in library/ gated by last_verified (5 production · 16 verified · 3 partial). New since v0.2.4: 4 modules (dashscope Alibaba CosyVoice TTS + voice cloning + Wanx, volcengine-tos S3-compat object storage, volcengine-speech ByteDance Seed-ASR 2.0 for QA-vs-script transcription diff, aitoearn 14-channel social publishing MCP) + library/cloudflare extended with the CF Pages source-repo silent-PATCH trap + delete+recreate runbook. Site now serves at trove.robozephyr.com (brand-aligned with the @robozephyr/ npm scope), with SVG favicon + Apple touch icon + auto-deploy workflow on every push. Format spec is stable; AI-assisted module authoring (v0.3) in progress.
See:
- SPEC.md — full format specification (Chinese, English translation forthcoming)
- ROADMAP.md — what's planned
- CONTRIBUTING.md — how to contribute modules
# 1. Install
npm install -g @robozephyr/trove
# (Node 22+ required. Homebrew tap planned for v1.0.)
# 2. Open the local Web UI to install a module and fill credentials
trove ui
# → http://127.0.0.1:7821
# Pick e.g. `minimax` from the Library tab, click Install, fill the form.
# 3. Reference it from any project's CLAUDE.md
echo '@/Users/you/.trove/minimax/module.md' >> /path/to/project/CLAUDE.md
# 4. Start your AI agent there
cd /path/to/project && claudeThe AI now knows MiniMax: how to call it, what gotchas to avoid, where to find the key.
If you don't want the binary, the directory convention IS the runtime — just clone, copy, edit:
mkdir -p ~/.trove
cp -r library/minimax ~/.trove/
chmod 600 ~/.trove/minimax/credentials.json
$EDITOR ~/.trove/minimax/credentials.json
# ⚠️ Direct file editing is the fallback. `trove ui` keeps secrets out of
# shell history / screenshots / pair-programming.
echo '@/Users/you/.trove/minimax/module.md' >> /path/to/project/CLAUDE.mdFor multi-account scenarios (e.g. multiple GitHub or Cloudflare accounts), duplicate the library/github-account/ template under a per-account directory name (e.g. ~/.trove/github-personal/, ~/.trove/github-work/) and fill each with that account's identity.
MIT