ToolMesh

The secure execution layer
for AI agent tool calls.

ToolMesh is an open-source middleware that sits between AI agents and your infrastructure. Authorization, credential security, audit trails, and content gating — in a single Go binary.

🤖 AI Agent

🔓 Auth OAuth 2.1 / API Key

🔐 AuthZ OpenFGA check

🔑 Creds Secret injected

🛡️ Gate PII redacted

⚡ Exec Tool executed

📋 Audit Logged to SQL

🌐 Backend

Get Started GitHub DADL Registry

Go binary Apache 2.0 licensed Self-hosted first

Nginx made web apps production-ready — reverse proxy, SSL, load balancing.
ToolMesh makes AI agent tool calls production-ready — authorization, credentials, audit, content gating.

Six Pillars

🔌

Any Backend

Connect existing MCP servers or describe REST APIs declaratively with DADL — no code required.

💻

Code Mode

LLMs write typed JavaScript instead of error-prone JSON. Saves context window budget.

📋

Audit

Every tool call is logged structurally and queryable via SQL. Full traceability.

🔐

Authorization

Fine-grained access control via OpenFGA. User → Plan → Tool relationships.

🔑

Credential Store

API keys and secrets are injected at runtime. The LLM never sees credentials.

🛡️

Output Gate

JavaScript policies validate inputs and filter outputs. PII redaction, compliance enforcement.

DADL — No More MCP Server Boilerplate

DADL (Dunkel API Description Language) describes any REST API as MCP tools in a single YAML file. No code, no server, no deployment.

mcp-server.js ~120 LOC

import { Server } from "@modelcontextprotocol/sdk";
import express from "express";

const app = express();
const server = new Server({ name: "github" });

server.setRequestHandler("tools/list", () => ({
  tools: [{
    name: "list_repos",
    description: "List repositories",
    inputSchema: {
      type: "object",
      properties: {
        sort: {
          type: "string",
          enum: ["created", "updated"]
        }
      }
    }
  }]
}));

server.setRequestHandler("tools/call",
  async (req) => {
    const resp = await fetch(
      "https://api.github.com/user/repos",
      { headers: {
          Authorization: "Bearer " + TOKEN
      }}
    );
    return { content: [
      { type: "text", text: await resp.text() }
    ]};
  });

app.use(server.transport);
app.listen(3000);
// + error handling, pagination,
// retries, auth refresh, types...

Also needed: Node.js runtime npm dependencies Docker image Process manager Health checks Deployment

→

github.dadl 15 LOC

spec: "https://dadl.ai/spec/v0.1"
backend:
  name: github
  type: rest
  base_url: https://api.github.com
  auth:
    type: bearer
    credential: github_token
  defaults:
    pagination:
      strategy: link_header
  tools:
    list_repos:
      method: GET
      path: /user/repos
      description: "List repositories"
      params:
        sort:
          type: string
          enum: [created, updated]

Auth, pagination, retries, error handling — all built in.

DADL files are generated by LLMs — ask your AI for a .dadl file and it works instantly. The first API description format that is AI-native created and AI-native consumed.

Caller-Origin — Know Who Is Calling

ToolMesh is the only known MCP gateway that differentiates which AI client triggers a tool call. CallerID + CallerClass enable tiered trust management.

CallerClass	PII Filtering	Tool Access
trusted	Credentials only	Full
standard	High-risk PII + credentials	Full
untrusted	All PII patterns	Sensitive tools blocked

Production-Ready Authentication

OAuth 2.1 + PKCE Standard-compliant interactive login

Multi-User Per-user plans, roles, and companies

API Keys Programmatic access with own identity

Simple Fallback Start with env vars, migrate to multi-user without architecture change

Get Started

ToolMesh is open source and self-hosted. Run it with Docker or as a Go binary.

Quick Start Guide Architecture