piLoci MCP is a self-hosted memory server for AI agents that exposes project-scoped memory storage and retrieval through the Model Context Protocol. Built to run on Raspberry Pi 5, it provides semantic recall, project listing, and user identity tools. Teams connect Claude Desktop, Codex, and other MCP clients to share persistent context without sending memory data to cloud services.
Use cases
- AI agent retrieves relevant project memories before starting a new task
- Team maintains shared context across multiple AI-assisted coding sessions
- Research agent searches historical project knowledge for relevant precedents
- Onboarding agent accesses project documentation and architecture decisions
- Cross-project agent queries only memories from authorized project scopes
Key features
- Claude Desktop
- Codex
- Cursor
Frequently Asked Questions
- What MCP tools does piLoci expose?
- The server exposes four main tools: memory (store new information), recall (semantic search), listProjects (enumerate accessible projects), and whoAmI (check current user identity).
- Does piLoci MCP require Raspberry Pi hardware?
- No, while optimized for Raspberry Pi 5, piLoci MCP runs on any Linux server with Node.js. It requires approximately 4GB RAM and 2GB storage.
- How does project isolation work?
- Each project has strict memory isolation. Users can only recall from projects they are authorized to access, preventing cross-project information leakage in multi-tenant team environments.
Related
Related
3 Indexed items
Mem0 MCP Server
Mem0 documents an official cloud-hosted Model Context Protocol server at docs.mem0.ai/platform/mem0-mcp, exposed over Streamable HTTP at `https://mcp.mem0.ai/mcp` with no local install required. Setup uses `npx mcp-add --name mem0-mcp --type http --url https://mcp.mem0.ai/mcp` for Claude, Claude Code, Cursor, Windsurf, VS Code, and OpenCode per Mem0 docs. Tools include `add_memory`, `search_memories`, `get_memories`, `get_memory`, `update_memory`, `delete_memory`, `delete_all_memories`, `delete_entities`, `list_entities`, `list_events`, and `get_event_status`. Authentication uses a Mem0 Platform API key (`MEM0_API_KEY`); the legacy mem0ai/mem0-mcp GitHub repo is archived in favor of this hosted server.
Graphiti MCP Server
Zep documents an experimental Graphiti Model Context Protocol server at help.getzep.com/graphiti/getting-started/mcp-server, with AI client integration also referenced at `https://help.getzep.com/_mcp/server` per Zep docs. The open-source implementation lives in getzep/graphiti/mcp_server with HTTP transport at `/mcp/` (default localhost:8000) or stdio mode. Tools include `add_episode`, `search_facts`, `search_nodes`, `get_episodes`, `delete_episode`, `clear_graph`, and `get_status` per the Graphiti MCP README; newer README entries also list `add_memory`, `add_triplet`, `search_memory_facts`, `summarize_saga`, `build_communities`, and `get_episode_entities`. Graphiti is Zep's open-source temporal knowledge graph framework for Context Graphs with real-time incremental updates and hybrid retrieval.
Mem0 MCP Server
Mem0 documents a hosted Model Context Protocol server at https://mcp.mem0.ai/mcp that exposes Platform memory tools (`add_memory`, `search_memories`, `get_memories`, `update_memory`, `delete_memory`, `delete_all_memories`, `delete_entities`, `list_entities`, `list_events`, `get_event_status`) to Claude, Claude Code, Codex, Cursor, Windsurf, VS Code, and OpenCode. Setup uses `npx mcp-add` with HTTP transport or manual JSON/TOML client configs; Codex requires `MEM0_API_KEY` as bearer token per docs.mem0.ai/platform/mem0-mcp. The cloud server needs a Mem0 Platform API key from the dashboard and Node.js for the installer—no local vector database required for the hosted path.