agentleFS
Frihet-ioFrihet-io/frihet-mcp/CLAUDE.md

frihet-mcp

CLAUDE.md13 starsChanged 5 days agoMITView on GitHub

Guidance for AI coding assistants working on this repository. See AGENTS.md for build commands and contribution conventions. MCP server that connects AI assistants (Claude · ChatGPT · Cursor · Windsurf · Cline · Antigravity · Codex · Copilot · Gemini CLI) to Frihet ERP. Natural language → invoices, expenses, clients, fiscal reports. Live: - npm: https://www.npmjs.com/package/@frihet/mcp-server — version: see package.json; surface counts are pinned by the generated public-capability contract - MCP remote: https://mcp.frihet.io (Cloudflare Worker) - Smithery: https://smithery.ai/servers/frihet/frihet-mcp - Anthropic…

The file

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

1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Every tool follows this contract:

    line 68

≈ “Common” is an estimate: at least 1 in 5 of the 26,980 CLAUDE.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
CLAUDE.md

CLAUDE.md — Frihet MCP Server

Guidance for AI coding assistants working on this repository. See AGENTS.md for build commands and contribution conventions.

What is this

MCP server that connects AI assistants (Claude · ChatGPT · Cursor · Windsurf · Cline · Antigravity · Codex · Copilot · Gemini CLI) to Frihet ERP. Natural language → invoices, expenses, clients, fiscal reports.

Live:

Repo: Frihet-io/frihet-mcp


Stack

  • Node.js >= 20
  • TypeScript (strict, target ES2022)
  • @modelcontextprotocol/sdk (peer dep)
  • Zero runtime deps (only 1 in package.json — minimal surface)
  • Distribution: npm + Cloudflare Worker (mcp.frihet.io) + Smithery
  • Tests: native node --test runner

Architecture

src/
  index.ts             — MCP server entry (Server + transport)
  client.ts            — Frihet API HTTP client (Bearer auth)
  client-interface.ts  — Typed interface for client mocking
  types.ts             — Shared TypeScript types
  logger.ts            — Structured logging
  observability.ts     — Langfuse LLM observability
  metrics.ts           — Tool call metrics
  openai-profile.ts    — OpenAI compatibility profile
  tools/
    register-all.ts    — authoritative tool registry (SoT — do not hand-count)
    <domain>.ts        — one module per domain (invoices, expenses, clients,
                         quotes, CRM, banking, fiscal/compliance, webhooks,
                         intelligence, …). `ls src/tools/` for the current list.
    shared.ts          — Cross-tool helpers
  resources/
    register-all.ts    — MCP resources (read-only context)
  prompts/
    register-all.ts    — MCP prompts (templated)

Cross-references

  • API client: hits https://api.frihet.io/v1 (managed in src/client.ts)
  • Auth: configured by the selected local or remote transport
  • Observability: optional and constrained by docs/observability.md
  • Remote host code lives under workers/remote-mcp/

Tool design pattern

Every tool follows this contract:

server.registerTool(
  'create_invoice',
  {
    title: 'Create invoice',
    description: 'Create a draft invoice for a client and return the structured invoice.',
    inputSchema: { /* Zod or JSON schema */ },
    annotations: {
      readOnlyHint: false,
      destructiveHint: false,
      idempotentHint: false,
      openWorldHint: false,
    },
  },
  async (input) => {
    // 1. Validate input (Zod)
    // 2. Call Frihet API V1 via client
    // 3. Return human-readable content plus schema-valid structuredContent
    return {
      content: [{ type: 'text', text: JSON.stringify(result) }],
      structuredContent: result,
    };
  }
);

Rules:

  • Tool name: preserve the established canonical snake_case convention (for example create_invoice)
  • description clear in 1 line, mentions input + output
  • inputSchema strictly typed
  • Output: MCP content plus outputSchema-valid structuredContent
  • Idempotency where possible (use Idempotency-Key HTTP header)
  • Errors: preserve actionable, sanitized API remediation via shared helpers

Build & Test

npm run build          # tsc → dist/
npm test               # npm run build && node --test dist/__tests__/*.test.js
npm start              # node dist/index.js (local stdio)

Surface changes must update the generated public-capability contract and pass the repository's existing CI gates. Release procedures are maintained outside this public contributor guide.


Quality bar

Tool errors propagate to user agents which act on the user's business data — treat every change accordingly.

  • Idempotency — src/client.ts mints an Idempotency-Key for EVERY mutating request (src/__tests__/idempotency-key-contract.test.ts pins it on the wire). Accepting a caller-supplied key as a tool input is a separate, per-tool step: today only create_credit_note does (src/tools/invoices.ts). Adding it to a tool means adding its test in the same diff.
  • Input validation — strict Zod schemas. Reject ambiguous input rather than infer.
  • Auth scope — tools must respect API key scope. No privilege escalation.
  • Rate limiting — client-side backoff on 429. Don't burn the user's quota.
  • PII — never log full request bodies. Mask NIF/IBAN/email in logs.
  • Side effects — destructive tools (delete, refund) need an explicit confirmation pattern.

Contact

Maintainer: Frihet (https://frihet.io) · support@frihet.io

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.