The trade is the usual one for a hub — you authorize Maton rather than each vendor directly, and in exchange the number of integrations you maintain stops growing with the number of apps. The agent toolkits matter if you are building rather than chatting: the same tool set shows up natively in LangChain, CrewAI, the Vercel AI SDK and the OpenAI Agents SDK.
Maton's hosted server at https://mcp.maton.ai, which fronts a unified API across 150+ business apps, plus the plugins and agent-framework toolkits that wire it up for you.
- A single endpoint covering more than 150 connected apps, so one integration replaces a per-app server for each of them
- Two ways in: OAuth, completed in the browser on first use, or an API key sent as a bearer token
- A local proxy for clients that would rather run something on the machine than talk to a remote endpoint
- Ready-made plugins for Claude Code, Cursor, Codex and OpenClaw that register the server for you
- Toolkits that expose the same tools inside agent frameworks: LangChain, the Vercel AI SDK, OpenAI and MCP in TypeScript; OpenAI Agents SDK, LangChain, CrewAI and Strands in Python
- Tool discovery at the endpoint, so the tool list each framework sees reflects what Maton currently supports
A Maton account. Point a client at https://mcp.maton.ai over streamable HTTP and finish OAuth in the browser, or run the local proxy from npm with npx -y @maton/mcp --api-key=YOUR_MATON_API_KEY. Each connected app also needs to be authorized on the Maton side before its tools do anything.
One command plus a key — npx -y @maton/mcp --api-key=YOUR_MATON_API_KEY, then supply credentials