# Toolmesh

> Toolmesh is an MCP server for developer tools. Self-hosted MCP gateway that turns REST APIs into tools with auth, policies and audit. It runs as a hosted remote server, so there is nothing to install locally.

- **Kind:** MCP server
- **Platforms:** self-hosted, API
- **Repository:** https://github.com/dunkelcloud/toolmesh
- **MCP Registry name:** io.github.DunkelCloud/toolmesh
- **Categories:** [developer tools](https://ai-powered.com/categories/developer-tools), [security](https://ai-powered.com/categories/security)
- **Listing updated:** 2026-11-18

## About

Think of it as a key vault and checkpoint between agents and your systems. Clients connect to one MCP endpoint, while ToolMesh keeps the API keys in its own environment, adds them at call time and checks with OpenFGA whether the caller may use that tool. Every call, including denied and failed ones, produces an audit entry, stored as log lines or in SQLite.

REST backends are described in DADL files, a few lines of YAML per endpoint, and the DADL registry offers descriptions for 36 APIs. Existing MCP servers over HTTP or stdio can sit behind the same endpoint, with tools named backend_tool. Optional JavaScript policies check parameters before a call and filter results afterwards, for example to redact personal data.

In Code Mode the model sees only two tools, discover_tools and execute_code, then writes short JavaScript to chain calls in one round trip. The project is at version 0.x and runs with Docker Compose alongside KeyDB; a public demo serves Hacker News APIs.

## Features

- **DADL REST wrapping:** Describe an API in YAML and expose it as MCP tools.
- **Server-held credentials:** Keys never appear in prompts or client config.
- **OpenFGA authorization:** Grants by user, role, backend, tool or access class.
- **Audit log:** One entry per call with caller, parameters and outcome.
- **Code Mode:** Two tools let the model script many calls at once.
- **JavaScript policies:** Pre-call checks and result filtering.

## Good for

- Give agents access to internal REST APIs without sharing keys
- Audit which tool an agent called and with what
- Restrict a team to read-only tools
- Put several MCP servers behind one endpoint

## Tools

- `discover_tools`: Looks up tool signatures in Code Mode.
- `execute_code`: Runs short JavaScript that makes several tool calls.

## FAQ

### How do I run it?

With Docker Compose; it starts ToolMesh and KeyDB and listens on port 8123 by default.

### What authenticates clients?

OAuth 2.1 or an API key.

### Is it stable?

The README says it is at version 0.x and minor releases can change configuration.

### Can I try it first?

Yes, a public demo at demo.toolmesh.io serves the Hacker News APIs.

### What are the best alternatives to Toolmesh?

The closest alternatives to Toolmesh are REA, Context7, Chrome DevTools MCP, Codebase Memory and GitHub.

## Where this comes from

- Description: written from the project's README.

## Alternatives

- [REA](https://ai-powered.com/mcp/rea): REA is an open-source MCP server for developer tools.
- [Context7](https://ai-powered.com/mcp/context7): Context7 is a freemium MCP server for developer tools.
- [Chrome DevTools MCP](https://ai-powered.com/mcp/chrome-devtools-mcp-2): Chrome DevTools MCP is an open-source MCP server for browser automation.
- [Codebase Memory](https://ai-powered.com/mcp/codebase-memory): Codebase Memory is an open-source MCP server for knowledge and memory.
- [GitHub](https://ai-powered.com/mcp/github-2): GitHub is an open-source MCP server for developer tools.
- [Serena MCP: the IDE for your agent](https://ai-powered.com/mcp/serena-mcp-the-ide-for-your-agent): Serena MCP: the IDE for your agent is an open-source MCP server for developer tools.
- [awesome-mcp-servers](https://ai-powered.com/mcp/awesome-mcp-servers-2): awesome-mcp-servers is an open-source MCP server for developer tools.
- [XERJ](https://ai-powered.com/mcp/xerj): XERJ is an open-source MCP server for knowledge and memory.

---
Source: https://ai-powered.com/mcp/toolmesh
