Logseq/Proj/LsqMCP/PRD

• Logseq MCP Project Summary and Requirements

  • Project Summary

    • Goal: Identify or create a JSON-RPC-based MCP server that supports structured read/write access to multiple local Logseq graphs.
    • Scope: Deep technical survey and prototype evaluation of candidate MCP servers or components (e.g., @logseq/nbb-logseq, logseq-query) that meet project requirements.

  • Product Requirements Document (PRD)

    • Functional Requirements

      • ✅ Must read from and write to Logseq graphs (blocks, pages, properties).
      • ✅ Must support simultaneous access to multiple local graphs.
      • ✅ Must expose a JSON-RPC API conforming to the Model Context Protocol (MCP).
      • ✅ Must support periodic polling (real-time subscriptions not required).
      • ✅ Must run on macOS.
      • ✅ Must interoperate with AI agents using structured formats (e.g., Markdown with frontmatter, JSON).
      • ✅ Must work via either:
        • Logseq's official/local HTTP API (if viable), or
        • a Node/npm-based method like @logseq/nbb-logseq or logseq-query.

    • Non-Requirements

      • ❌ No need for authentication/authorization or multi-user permissioning.
      • ❌ No requirement for background/daemon mode (CLI or persistent server both acceptable).
      • ❌ No need for cross-platform support (only macOS required).
      • ❌ No constraint on programming language or runtime.

    • Open Questions

      • ❓ What are the current capabilities and limitations of logseq-query and @logseq/nbb-logseq regarding write operations?
      • ❓ Can the MCP server dynamically discover and reload graphs or must graphs be statically configured?