MCP server to interact with LogSeq via its API.

The server implements multiple tools to interact with LogSeq:

• list_graphs: Lists all available graphs
• list_pages: Lists all pages in the current graph
• get_page_content: Return the content of a single page
• search: Search for content across all pages
• create_page: Create a new page
• update_page: Update content of an existing page
• delete_page: Delete a page

It's good to first instruct Claude to use LogSeq. Then it will always call the tool.

Example prompts:

• Get the contents of my latest meeting notes and summarize them
• Search for all pages where Project X is mentioned and explain the context
• Create a new page with today's meeting notes
• Update the project status page with the latest updates

You can configure the environment with LogSeq API settings in two ways:

1. Add to server config (preferred)

{
  "mcp-logseq": {
    "command": "uvx",
    "args": [
      "mcp-logseq"
    ],
    "env": {
      "LOGSEQ_API_TOKEN": "<your_api_token_here>",
      "LOGSEQ_API_URL": "http://localhost:12315"
    }
  }
}

2. Create a .env file in the working directory with the required variables:

LOGSEQ_API_TOKEN=your_token_here
LOGSEQ_API_URL=http://localhost:12315

To prepare the package for distribution:

1. Sync dependencies and update lockfile:

uv sync

Since MCP servers run over stdio, debugging can be challenging. For the best debugging experience, we recommend using the MCP Inspector.

You can launch the MCP Inspector via npm with:

npx @modelcontextprotocol/inspector uv --directory /path/to/mcp-logseq run mcp-logseq