agentleFS
adezeadeze/raindrop-mcp/AGENTS.md

raindrop-mcp

AGENTS.md188 starsChanged 8 months agoMITView on GitHub

TypeScript 7/Bun MCP server for Raindrop.io bookmarks, using Vite+ 1.1. This is the canonical operating guide; CLAUDE.md and .github/copilot-instructions.md are compatibility pointers. Target MCP SDK v2 and protocol revision 2026-07-28. Run the irreversible live bookmark test only with an account authorized for it: It creates a uniquely tagged project URL, retrieves it by its returned ID, deletes it from the active library and Trash, then confirms it is absent. It never targets an existing bookmark. Use Bun for project commands…

  • Reads credentials

The file

We split this AGENTS.md into the rules it gives an agent

8 rules. About 0 are common, 2 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Run the irreversible live bookmark test only with an account authorized for it:

    line 17
  2. Very specific:

    Use Bun for project commands and dependency updates: bun run update:deps.

    line 27
  3. Very specific:

    Add tools through domain modules with defineTool().

    line 41
  4. Very specific:

    Prefer mcp://collection/{id} and mcp://raindrop/{id} resource content for list/detail links.

    line 43
  5. Very specific:

    Use createLogger() from src/utils/logger.ts.

    line 44
  6. Very specific:

    Treat raindrop-complete.yaml as the Raindrop API contract.

    line 45
  7. Very specific:

    Use @modelcontextprotocol/server, @modelcontextprotocol/client, and @modelcontextprotocol/node.

    line 46
  8. Distinctive:

    Update README.md for user-visible commands/features.

    line 52

≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
AGENTS.md

Raindrop MCP — Codex Guide

TypeScript 7/Bun MCP server for Raindrop.io bookmarks, using Vite+ 1.1. This is the canonical operating guide; CLAUDE.md and .github/copilot-instructions.md are compatibility pointers. Target MCP SDK v2 and protocol revision 2026-07-28.

Commands

bun install --frozen-lockfile
bun run check
bun run lint
bun run format:check
bun run type-check
bun run test
bun run build

Run the irreversible live bookmark test only with an account authorized for it:

RUN_DESTRUCTIVE_LIVE_TESTS=true \
CONFIRM_DESTRUCTIVE_LIVE_TESTS=DELETE_ONLY_TEST_BOOKMARKS \
bunx vp test run tests/destructive.live.test.ts

It creates a uniquely tagged project URL, retrieves it by its returned ID, deletes it from the active library and Trash, then confirms it is absent. It never targets an existing bookmark.

Use Bun for project commands and dependency updates: bun run update:deps. Keep bun.lock committed. Do not manually version, tag, publish, or edit release artifacts; semantic-release owns releases on master.

Use Node 24.15+ and Bun 1.4.3+ for development; the published server requires Node 22.19+. vite.config.ts owns lint, formatting, tests, coverage, and vp pack configuration. Keep Vitest and coverage aligned with Vite+'s bundled version. tooling/compiler-api isolates TypeDoc's required TypeScript 6 compiler; the server uses TypeScript 7. Run bun run docs from the root.

Architecture

  • src/index.ts: v2 STDIO era-selection entry point; src/server.ts: v2 per-request Streamable HTTP handler.
  • src/services/raindropmcp.service.ts: MCP capability, tool, resource, and prompt registration. Register capabilities before handlers.
  • src/tools/*.ts: declarative tool modules, assembled by src/tools/index.ts.
  • src/services/raindrop.service.ts: shared authenticated Raindrop API client, rate limit, and typed upstream errors.
  • tests/: Vitest unit/integration coverage; E2E needs RAINDROP_ACCESS_TOKEN.

Non-negotiable contracts

  • Add tools through domain modules with defineTool(); use snake_case, Zod input/output schemas, async handlers, and the shared RaindropService.
  • Destructive operations require confirm: true; preserve existing auth, HTTP security, rate-limit, and typed-error behavior.
  • Prefer mcp://collection/{id} and mcp://raindrop/{id} resource content for list/detail links.
  • Use createLogger() from src/utils/logger.ts; never write protocol output with console.log.
  • Treat raindrop-complete.yaml as the Raindrop API contract. Regenerate types/client only when its OpenAPI input changes.
  • Use @modelcontextprotocol/server, @modelcontextprotocol/client, and @modelcontextprotocol/node; do not add the v1 monolithic SDK.

Codex workflow

  • Inspect current implementation and tests before changing protocol behavior. Use official MCP SDK/Raindrop documentation when an API contract is uncertain.
  • Preserve unrelated working-tree changes. Keep diffs scoped; run focused tests while iterating, then lint, format:check, type-check, test, and build for cross-cutting changes.
  • Update README.md for user-visible commands/features; update this file when architecture or contributor workflow changes.
  • VS Code files and Copilot-era guidance are optional compatibility material, not the project control plane.

Release

CI uses Bun plus Node 24 and frozen installs. Merge conventional commits to master; CI runs semantic-release and synchronizes package.json, manifest.json, mcp.json, gemini-extension.json, changelog, and the MCPB bundle.

.github/workflows/publish.yml is the npm-trusted workflow. Its manual dry_run input defaults to true; disable it only for authorized publication. Release jobs are serialized and use OIDC without NPM_TOKEN. Protocol integration tests mock upstream responses and use a dummy Raindrop token.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.