Lamatic MCP
Build, deploy, run and debug your Lamatic flows from any AI assistant.
Lamatic MCP (formerly Graph MCP) is a hosted Model Context Protocol server at https://mcp.lamatic.ai/mcp. Connect Claude, Cursor, VS Code or any other MCP client and your assistant can list your projects, create and edit flows, deploy them, run them, read logs, and work with your D1 database and VectorDB collections.
There is nothing to install and no API key to copy. You sign in with your Lamatic account in the browser (OAuth 2.1 with PKCE), and the client keeps its own short-lived tokens that refresh automatically.
Using @lamatic/graph-mcp or @lamatic/dev-mcp from npm? Those packages, and the graph_auth_login / dev_auth_login tools that took an org API key (lt-org-…) and project API key (lt-mmup-…), are deprecated. Remove them from your MCP config and connect to https://mcp.lamatic.ai/mcp instead.
Connect
Claude Code
claude mcp add --transport http lamatic https://mcp.lamatic.ai/mcpThen run /mcp, pick lamatic, and choose Authenticate.
Claude (claude.ai and Claude Desktop)
Go to Settings → Connectors → Add custom connector and enter https://mcp.lamatic.ai/mcp. Claude asks you to sign in the first time you use it.
Cursor
Use the Add to Cursor button above, or add this to your MCP config file:
{
"mcpServers": {
"lamatic": {
"url": "https://mcp.lamatic.ai/mcp"
}
}
}VS Code
Use the Add to VS Code button above, or add this to .vscode/mcp.json:
{
"servers": {
"lamatic": {
"type": "http",
"url": "https://mcp.lamatic.ai/mcp"
}
}
}Other clients
Any client that supports remote (Streamable HTTP) MCP servers with OAuth works. Add https://mcp.lamatic.ai/mcp as the server URL.
Signing in and permissions
The first time your client connects, it opens a Lamatic sign-in page in your browser. There you choose:
- The organization this connection works in. One connection reaches one organization.
- Which projects: all projects in that organization, or only the ones you pick.
- What it may do: a preset (full access, builder, or read-only) or individual permissions. You can never grant more than your own role in that organization allows.
- An optional expiry for the connection.
No token passes through the chat, and nothing is stored on the MCP server between requests. If your role in the organization is reduced, the connection loses those permissions right away.
To see, change or disconnect your connected assistants, open Connected apps in Lamatic Studio. Disconnecting takes effect within about 5 minutes for an assistant that is in the middle of a session.
To work in a different organization, ask your assistant to run graph_switch_organization (or connect again), and pick the other organization on the sign-in page.
Ask your assistant to run graph_auth_status at any time to see which organization, role and permissions the connection has, and which tools it can use.
Tools
Every project tool takes a projectId. Your assistant gets it from graph_list_projects, so you can refer to projects by name.
Account and projects
| Tool | Description |
|---|---|
graph_auth_status | Show the organization, role and permissions of this connection, and which tools it can use |
graph_list_projects | List the projects this connection can work in |
graph_get_project | Read one project’s name, organization and settings |
graph_switch_organization | Reconnect to a different organization (signs in again) |
graph_get_guide | Read a built-in guide: capabilities, conventions, troubleshooting, flow-template-chat |
Flows
| Tool | Description |
|---|---|
graph_list_flows | List active flows with their input schemas |
graph_load_flow_details | Load one flow’s details and input schema |
graph_create_flow | Create a flow from a node and edge graph |
graph_update_flow | Replace a flow’s nodes and edges and/or its status (a full replace, not a merge) |
graph_execute_flow | Run a deployed flow with a payload |
graph_generate_flow_skills | Generate one Claude Skill per flow (what it does, when to use it, how to call it) |
Nodes and models
| Tool | Description |
|---|---|
graph_list_nodes | List the node types available for building flows |
graph_get_node | Get one node type’s full configuration schema |
graph_list_models | List the model credentials configured in a project |
graph_list_model_providers | List supported model providers and the fields each one needs |
graph_get_model_status | Check whether a model is currently available |
graph_create_model_credential | Add a model provider credential. You enter the key on a Lamatic page, never in the chat |
Deployments
Flow changes are not live until you deploy.
| Tool | Description |
|---|---|
graph_trigger_deployment | Deploy the project’s current flows (runs in the background) |
graph_get_deployment | Get one deployment’s status, what changed and how long it took |
graph_list_deployments | List the project’s deployments |
Logs
| Tool | Description |
|---|---|
graph_get_logs | Fetch execution logs for a project |
graph_enable_logs | Turn execution logging on (takes effect after the next deployment) |
graph_disable_logs | Turn execution logging off (takes effect after the next deployment) |
D1 database
| Tool | Description |
|---|---|
graph_d1_list_tables | List the tables in the project’s D1 database |
graph_d1_get_table_columns | Get a table’s columns |
graph_d1_create_table | Create a table |
graph_d1_drop_table | Delete a table |
graph_d1_execute_query | Run a SQL query |
VectorDB
| Tool | Description |
|---|---|
graph_vectordb_list_collections | List collections |
graph_vectordb_get_collection | Get a collection’s properties and record count |
graph_vectordb_create_collection | Create a collection |
graph_vectordb_delete_collection | Delete a collection |
graph_vectordb_list_records | List records in a collection |
graph_vectordb_get_record | Get one record by ID |
graph_vectordb_search | Search a collection, ranked by relevance |
graph_vectordb_update_record | Update a record’s fields |
The tools you see depend on the permissions you approved. A read-only connection, for example, does not get the tools that change anything.
Prompts
Clients that support MCP prompts show these as slash commands:
/build_flow: build a new flow step by step/debug_flow: find out why a flow fails/deploy_and_verify: deploy and confirm the change is live
Supported flow triggers
graph_execute_flow can run flows with these triggers:
| Trigger | Supported |
|---|---|
| API Request | âś… |
| Chat Widget | âś… |
| Search Widget | âś… |
| Webhook | âś… |
| Cron | ⚠️ Needs trigger data in the payload |
| Google Sheets | ⚠️ Needs trigger data in the payload |
Running a flow needs an active project API key in Studio. Lamatic uses it on the server for you, so it never reaches your assistant.
Example prompts
list my Lamatic projectsshow me the flows in the Support projectrun the Changelog flow with sampleInput "hello world"add a step to the RAG flow that summarizes the answer, then deploy itwhy did the last run of the Onboarding flow fail?Links
- Connected apps in Lamatic Studio
- GitHub repository
- Docs MCP: ask questions about Lamatic from your AI assistant
- MCP Node: call external MCP servers from inside a flow