Locus
Give coding agents a structured markdown “memory palace” they can navigate so long-running projects keep context organized instead of one flat notes file.
Overview
Locus is a MCP server for the Build phase that exposes a hierarchical markdown memory palace and navigation tools so AI agents can structure and recall long-term project memory.
What is this MCP server?
- Hierarchical markdown memory palace designed for AI agent navigation
- MCP tools for structured palace traversal and recall
- PyPI package locus-mcp 0.9.0 with stdio transport
- Local-first palace storage in markdown hierarchy
- Repository: Nano-Nimbus/locus on GitHub
- Package version: locus-mcp 0.9.0 (PyPI, stdio)
- MCP registry server version: 0.9.0
What problem does it solve?
Long agent sessions lose track of where facts live because context is flat, duplicated, or buried in chat history.
Who is it for?
Indie builders running weeks-long agent projects who want durable, navigable memory separate from ephemeral chat.
Skip if: Teams that only need semantic search over embeddings and do not want to maintain a palace hierarchy by hand.
What do I get? / Deliverables
Your agent can walk a structured markdown palace via MCP and pull the right room of memory without re-scraping the whole repo every turn.
- MCP-navigable hierarchical memory palace
- Structured agent recall across coding sessions
- Palace nodes maintained as markdown hierarchy
Recommended MCP Servers
Journey fit
Agent memory and palace navigation are core build-phase agent tooling for anyone shipping products with Claude Code, Cursor, or similar MCP clients. Agent-tooling subphase is the canonical shelf for persistent hierarchical memory exposed as MCP tools, not one-off chat context.
How it compares
MCP memory-palace navigation over markdown, not a RAG vector DB MCP or a single-note skill.
Common Questions / FAQ
Who is Locus for?
Solo builders and small teams using MCP agents who want hierarchical, agent-navigable project memory stored as markdown.
When should I use Locus?
Use it while building agent-driven products when context spans many sessions and you need structured recall beyond the default context window.
How do I add Locus to my agent?
Install the locus-mcp PyPI package (0.9.0), configure stdio MCP in your client, and point it at your local palace markdown tree per the Locus repo docs.