Mnemo - Persistent AI Memory

Instructions

Implementation Reference

• src/mnemo_mcp/server.py:671-687 (handler)

  Handler function that loads and returns documentation for specified topics (memory or config). Validates topic parameter against whitelist, loads markdown file from mnemo_mcp.docs package using importlib.resources, and returns content or error JSON.

  async def help(topic: str = "memory") -> str:
      """Load full documentation for a tool."""
      docs_package = pkg_resources.files("mnemo_mcp.docs")
      valid_topics = {"memory": "memory.md", "config": "config.md"}
  
      filename = valid_topics.get(topic)
      if not filename:
          return _json(
              {
                  "error": f"Unknown topic: {topic}",
                  "valid_topics": list(valid_topics.keys()),
              }
          )
  
      doc_file = docs_package / filename
      content = doc_file.read_text(encoding="utf-8")
      return content

• src/mnemo_mcp/server.py:661-670 (registration)

  Registration of the 'help' tool with FastMCP decorator. Defines tool metadata including description, title, and behavior hints (read-only, non-destructive, idempotent).

  @mcp.tool(
      description="Full documentation for memory and config tools. topic: 'memory' | 'config'",
      annotations=ToolAnnotations(
          title="Help",
          readOnlyHint=True,
          destructiveHint=False,
          idempotentHint=True,
          openWorldHint=False,
      ),
  )

• src/mnemo_mcp/server.py:671-671 (schema)

  Schema definition: accepts 'topic' parameter (default 'memory'), validated against valid_topics dict. Returns string (either markdown content or JSON error).

  async def help(topic: str = "memory") -> str: