AI AGENT × MIND MAP × MCP
Let AI agents create and edit visual mind maps directly
GoMind MCP connects your local mind map canvas to compatible AI agents such as Codex, Claude Code, and Visual Studio Code. Ask an agent to read the current structure, propose a plan, create new branches, or refine existing content in natural language. The result remains an editable mind map in GoMind.
What is GoMind MCP?
GoMind provides an MCP Server on your local computer, while your AI agent acts as the MCP Host. Once connected, the agent can use the canvas tools exposed by GoMind for reading, planning, and editing. The connection is local-first: it does not require a separate GoMind MCP membership, and any subscription or API cost belongs to the agent or model provider you choose.
MCP and the GoMind Skills API
Normal AI agents should use standard MCP and the /mcp endpoint. Use /gomind_openclaw/manifest and /gomind_openclaw/invoke only when OpenClaw or another tool explicitly asks for GoMind's dedicated Skills API. Both paths support AI-driven mind map workflows, but their protocols and setup entries are different.
What can GoMind MCP help you do?
- Understand the current canvas: summarize the central topic, main branches, and existing structure before taking action.
- Turn a description into structure: convert a product plan, study outline, meeting note, or research brief into a mind map.
- Keep collaborating: ask for a proposal first, confirm it, and then write it to GoMind without losing existing content.
- Connect your workflow: use standard MCP Streamable HTTP with compatible clients and discover the tools available in your current GoMind version.
Connect GoMind MCP in three steps
- Start the local service in GoMind. Open Settings → AI & Integrations → Local AI Connection, choose Start connection, select MCP, and copy the complete endpoint shown by the setup wizard. It usually looks like:
http://127.0.0.1:<actual-port>/mcp
- Add the server to your AI agent. Choose HTTP or Streamable HTTP and paste the endpoint copied from GoMind.
- Read first, then write. Start with a read-only request to confirm tool discovery. Then ask for a bounded structure proposal before creating or changing content.
Codex CLI or configuration file
[mcp_servers.gomind]
url = "http://127.0.0.1:<actual-port>/mcp"
Claude Code
claude mcp add --transport http gomind \
--scope user "http://127.0.0.1:<actual-port>/mcp"
Run claude mcp list to check the server.
Visual Studio Code
{
"servers": {
"gomind": {
"type": "http",
"url": "http://127.0.0.1:<actual-port>/mcp"
}
}
}
Use MCP: List Servers to start or restart GoMind.
Prompts that work
When connecting for the first time, verify that the agent can read the canvas before asking it to write:
Read the currently open mind map through GoMind and summarize its central topic and main branches. Do not modify anything yet.
Once the connection works, give the agent a bounded task:
First read the current mind map and understand its structure. Organize a "New Product Launch Plan" into five first-level branches: Market Research, Product Readiness, Channel Promotion, Launch Execution, and Retrospective. Add 3 to 5 concrete actions under each branch. Show me the structure proposal first and wait for my confirmation before writing it to GoMind. Keep existing content and do not delete branches.
Data and permission boundaries
- Current canvas: reading and editing the currently open mind map does not require permission for your whole disk or a folder.
- Authorized folders: only add a folder when the task needs files from it, and choose the smallest useful scope.
- Local connection: the endpoint binds to
127.0.0.1for clients on the same computer. Do not use port forwarding or a public tunnel. - Model providers: MCP runs locally, but content you provide to an agent may still be sent to its model provider. Follow that provider's privacy policy for sensitive material.
Frequently asked questions
What is GoMind MCP?
It is a standard MCP Server provided locally by GoMind. A compatible AI agent can use Streamable HTTP to read the current canvas and create or adjust an editable mind map.
Where do I copy the endpoint?
Start MCP under Settings → AI & Integrations → Local AI Connection. The completed connection screen shows the actual endpoint; its port can vary if another application is using a port.
Why is the client connected but showing no tools?
Make sure GoMind is still running and the local service has not been stopped. Refresh or restart MCP in the client. A running service in GoMind does not by itself prove that the client has completed tool discovery.
Can a cloud agent or another computer connect?
Not by default. 127.0.0.1 points to the computer running the client, and GoMind's local MCP service is normally intended for the same computer where GoMind is installed.
How do I disconnect?
Disable or remove the GoMind server in your AI client, or stop the local service from GoMind's Local AI Connection management page.
Continue with GoMind
Open GoMind, start the local MCP service, and turn your next brainstorm, product plan, or research session into a map you can keep working on.
FAQ
Frequently asked questions
What is GoMind MCP?
It is a standard MCP Server provided locally by GoMind. A compatible AI agent can use Streamable HTTP to read the current canvas and create or adjust an editable mind map.
Where do I copy the endpoint?
Start MCP under Settings, AI & Integrations, and Local AI Connection. The completed connection screen shows the actual endpoint; its port can vary if another application is using a port.
Why is the client connected but showing no tools?
Make sure GoMind is still running and the local service has not been stopped. Refresh or restart MCP in the client. A running service does not by itself prove tool discovery is complete.
Can a cloud agent or another computer connect?
Not by default. 127.0.0.1 points to the computer running the client, and GoMind’s local MCP service is normally intended for the same computer where GoMind is installed.
How do I disconnect?
Disable or remove the GoMind server in your AI client, or stop the local service from GoMind’s Local AI Connection management page.
KEEP BUILDING
Let AI organize the work. Keep the final structure yours.
If you are new to local AI connections, start with the AI and integrations section of the GoMind user guide.
Read AI and integrations guide →