KnowMan MCP server for Jan

What you need first

Jan supports the Model Context Protocol (MCP), an open standard for letting language models interact with external tools and data sources, and works as an MCP host (Jan docs). For background, see the KnowMan MCP server guide.

You need three things. Node.js installed on your machine, because Jan requires NodeJS and/or Python for MCP servers (Jan docs), and the KnowMan package needs Node 22 or newer (npm). A KnowMan account at app.knowman.io. And a KnowMan API key, or the one-time browser sign-in described below.

Create a KnowMan API key

Open app.knowman.io and go to Settings > Connections > Connect an AI agent. Create a key there. Keys look like km_... (GitHub).

Add the server in Jan

MCP servers in Jan are configured under Settings → MCP Servers (Jan docs). Click the + sign in the upper right-hand corner of the MCP box, then fill in the form with the values shown above (Jan docs). A server entry consists of Server Name, Command, Arguments, and Environment Variables fields (Jan docs).

Jan's documented setup runs local servers through a command, such as npx with a package argument (Jan docs). @knowman/mcp is KnowMan's stdio server package, and KNOWMAN_API_KEY is the environment variable it reads (npm).

Or sign in once with knowman login

If you would rather not paste a key into the form, run npx -p @knowman/mcp knowman login once in a terminal. It opens your browser for sign-in and saves the credential (npm).

The saved credential is shared by the CLI and the MCP server, so you can leave the Environment Variables field empty (npm).

Check that it works

After saving the entry, check that the server has been activated successfully (Jan docs). Then ask Jan to use it, for example "search my KnowMan for ..." or "save this URL" (GitHub).

By default you must approve each tool call individually, and approval panels appear inline in the chat, showing the exact arguments before you accept or deny (Jan docs). To pre-approve everything, enable Allow All MCP Tool Permission in Settings (Jan docs).

The six KnowMan tools

The server exposes six tools (npm). kb_search searches saved items by keyword or meaning and returns titles, URLs and metadata. kb_ask answers questions with citations. kb_get_item returns the readable content of one item. kb_save saves a URL to the Inbox or the Knowledge Base with an optional note. kb_recent lists recent items. kb_memories lists saved claims, decisions and playbooks.

Pick a model that can call tools

The model you use must have tools enabled for MCP to work (Jan docs). For local models, enable tool calling in the model parameters with the edit button in Model Capabilities (Jan docs).

Not all downloadable models are good at tool calling or compatible with MCP, and many open-source models are not designed for tool use, so you may need to try a different model (Jan docs). Jan's docs recommend a model with strong tool-use capabilities, such as Claude Sonnet or Opus (Jan docs). If you enter an Anthropic API key under Model Providers > Anthropic, Jan lists the available Anthropic models automatically (Jan docs).

Troubleshooting and cautions

If the server will not connect, make sure you typed the commands correctly in the MCP Server form, restart Jan, and make sure the model you are using has tools enabled (Jan docs). If the key might be wrong, create a new one under Settings > Connections > Connect an AI agent in the KnowMan web app (GitHub).

Two cautions from Jan's docs. Active MCP connections may consume part of a model's context window, which could affect performance. Granting a model access to external tools is a significant security consideration, including prompt injection risks (Jan docs).