Awesome MCP Registry

What is Awesome MCP Registry?

Awesome MCP Registry is an enterprise MCP Registry & Gateway for discovering, securing, and executing Model Context Protocol (MCP) tools. It unifies access to local STDIO processes and remote StreamableHTTP servers via a single HTTP API, with built-in Sandbox and OAuth 2.1 support. It currently indexes 4547+ MCP servers.

How to use Awesome MCP Registry?

Deploy the full gateway with docker compose up -d (recommended), or use the NPM package as a data-only SDK, or fetch the public JSON endpoint directly. For tool execution, send HTTP POST requests to /api/v1/packages/run or use the Streamable HTTP proxy at /mcp/<packageName>. Contributors can submit new MCP servers via the Contributing Guide.

Key features of Awesome MCP Registry

• Federated registry searching local and official @modelcontextprotocol/registry servers
• Unified HTTP API for local STDIO and remote StreamableHTTP tools
• Secure sandboxed execution (supports E2B, Daytona, Sandock)
• Built-in OAuth 2.1 proxy for authentication flows
• Private, self-hosted Docker deployment
• Auto-generated OpenAPI/Swagger documentation

Use cases of Awesome MCP Registry

• Centralize tool access for enterprise LLM applications
• Run community MCP servers safely in isolated sandboxes
• Bridge local CLI tools to remote AI agents via HTTP API
• One API to search and manage thousands of MCP tools

FAQ from Awesome MCP Registry

What does validated: true mean?

Packages marked validated can be loaded via ToolSDK client to get ready-to-use adapters (e.g., Vercel AI SDK tools) without writing custom schema mapping.

How do I deploy the Awesome MCP Registry?

Run docker compose up -d. Configure via .env (e.g., set MCP_SANDBOX_PROVIDER=LOCAL to disable the sandbox). The registry starts with a web UI, Swagger docs, and full-text search.

Can I access the registry data without self-hosting?

Yes. Use the NPM package @toolsdk.ai/registry or fetch the public JSON index at https://toolsdk-ai.github.io/toolsdk-mcp-registry/indexes/packages-list.json.

How do I execute a tool remotely via HTTP API?

Send a POST request to /api/v1/packages/run with the package name, tool key, and input data. For Streamable HTTP access, use /mcp/<packageName> and pass environment variables via x-mcp-env-* headers.

How do I submit my own MCP server to the registry?

Follow the Contributing Guide in the repository, which details the JSON schema, examples for remote servers and OAuth 2.1, and the submission process.