Summary
Your notes graph, writable from the conversation.
Read-only note access is common; the block-level writes are what make this different. An assistant can insert a child block under the right parent instead of appending text to the bottom of a page.
What it is
An MCP server that talks to a running Logseq instance over its local API. Tools sit under a logseq namespace and split into pages and blocks, which is how Logseq itself is structured.
What you get
- Every page in the graph listed, and any single page fetched
- Pages created, and deleted along with their blocks
- All blocks on a page read, or one block by ID
- Blocks created on a page, or inserted as a child of another block
- Blocks updated, moved and removed with their children
- Blocks searched by query across the whole graph
Requirements
A recent Python, and Logseq with Developer mode on, the Logseq Developer Plugin enabled and an API token set in Advanced settings — then restart Logseq. The server reads LOGSEQ_API_URL, http://localhost:12315 in the standard setup, and LOGSEQ_TOKEN.
Setup effort
Build from source — clone the repository and build it, then point your client at the binary