MCP in Lamatic
Model Context Protocol (MCP) is an open standard for connecting AI assistants to external systems, data sources, APIs, tools, and services through one uniform interface.
Lamatic ships three MCP products, each covering a different direction of the protocol:
- MCP Node lets your flow consume tools from any external MCP server.
- Graph MCP exposes your deployed flows as MCP tools that external AI assistants can call.
- Docs MCP is a hosted MCP server that answers questions about Lamatic itself.
The three MCPs
MCP Node: consume external MCP servers
The MCP Node connects your Lamatic flow to an existing MCP server. Configure the server’s URL, headers, and protocol once, and any AI node downstream can call the tools that server exposes.
Use it when you want your flow to call a third-party MCP server, such as a weather API, a database, or an internal tool catalog.
Graph MCP: expose your flows as MCP tools
Graph MCP runs in the opposite direction: it turns your deployed Lamatic flows into MCP-callable tools. Any MCP client (Claude Desktop, Cursor, VS Code, custom agents) can then invoke those flows with natural language.
Use it when you want an external AI assistant to drive your flows. For example, a developer in Cursor triggering a flow that generates a weekly report.
Docs MCP: query Lamatic docs from any AI assistant
Docs MCP is a hosted MCP server that Lamatic maintains. It indexes lamatic.ai/docs into a RAG pipeline and exposes a query_docs tool. Any MCP client can ask natural-language questions about Lamatic and get answers with references. No setup, no API key.
Use it when you want your AI assistant to answer questions about Lamatic itself its nodes, agents, integrations and error messages, instead of you searching the docs manually.
When to use which
| MCP Node | Graph MCP | Docs MCP | |
|---|---|---|---|
| Direction | Lamatic flow → external MCP server | External MCP client → Lamatic flow | External MCP client → Lamatic-hosted docs server |
| Role | Client (consumer) | Server (provider) | Server (hosted by Lamatic) |
| What you provide | The URL of an external MCP server | A deployed flow with an API trigger | Nothing — it’s hosted |
| What you get | External tools available inside the flow | Your flow callable from any MCP client | A query_docs tool that answers questions about Lamatic |
| Typical caller | AI nodes inside your flow | Claude Desktop, Cursor, VS Code, custom agents | Claude Desktop, Cursor, anyone reading the docs |
| Auth model | Headers and API keys you set on the node | Org API key + Project API key | None (free to use) |
| Setup location | Studio → MCP/Tools → MCP | npx @lamatic/graph-mcp in the client’s config | Add https://docmcp.lamatic.ai/api/mcp to your client config |
Quick chooser
- “I want my flow to use tools from an external service via MCP.” → MCP Node
- “I want Claude or Cursor to call my Lamatic flow.” → Graph MCP
- “I want my AI assistant to answer questions about Lamatic itself.” → Docs MCP
- All of the above. → A flow that consumes external MCP tools (MCP Node) can itself be exposed via Graph MCP, while you use Docs MCP to ask your assistant about Lamatic as you build.