Skip to main content
MCP server CX Codex

MemoryPilot — Claude Code MCP server: what it does, how to install

MemoryPilot is an MCP server for Claude Code that stores context locally across sessions without third-party API calls.

no license ★ 18 updated 2026-10-29

Open on GitHub

What MemoryPilot does

MemoryPilot tackles the issue of AI coding assistants losing memory between sessions. It combines BM25 keyword matching with jina-embeddings-v5-nano retrieval and a cross-encoder reranker. The engine also builds a temporal knowledge graph, tagging entries with importance ratings and expiration limits directly upon insertion.

Everything runs inside a single compiled Rust binary with zero runtime dependencies and no remote API calls. Built-in AAAK compression shrinks token usage by 5 to 10 times during context injection. Cross-encoder reranking takes roughly 900 ms per query on Apple silicon, but setting MEMORYPILOT_CROSS_RERANK=off drops latency down to 28 ms.

What it helps with

  • retain project knowledge across separate assistant sessions
  • cut prompt token usage 5 to 10 times via AAAK compression
  • retrieve relevant notes locally through hybrid search without API calls
  • clean up obsolete data automatically via database self-compaction

Example prompts for the agent

  • Recall the architectural decisions we agreed on in earlier sessions for this project.
  • Search long-term memory for our database configuration settings.

How to install MemoryPilot

Install

The MCP registry has no launch command for it: see the repository README

Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more

Questions about MemoryPilot

How to install MemoryPilot in Claude Code?

The tool is distributed as a standalone Rust binary. Download the latest release binary from the project repository and register its executable path in your MCP client settings.

What does MemoryPilot do?

It gives AI assistants persistent local memory across sessions. It performs hybrid search, compresses retrieved context to save tokens, and structures facts with a knowledge graph.

Does MemoryPilot require external API keys for embeddings?

No, it makes zero API calls. Embedding generation and reranking run entirely on your local machine inside the binary.

Similar MCP servers

All MCP servers for Agent setup: memory, context, orchestration →
memory-lancedb-pro win4r A Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, and OpenCode skill. It covers the memory-lancedb-pro plugin architecture: retrieval pipeline, scopes, tools, and CLI. CX Codex +1 647installs sqlite-graph-memory tonydzi sqlite-graph-memory gives coding agents associative long-term memory over markdown notes. It pairs dense vector retrieval with existing [[wikilinks]] using SQLite and zero external graph databases. CX Codex +1 ★ 17repo stars claude-code-plugin-for-goodmem-a-memory-infra-for-agents PAIR-Systems-Inc claude-code-plugin-for-goodmem-a-memory-infra-for-agents is a Claude Code plugin that connects to GoodMem memory infrastructure for managing embedders, rerankers, and memory spaces. It helps build RAG pipelines and research agents. ★ 10repo stars ai-data-engineering-rag-pipeline reason-machines ai-data-engineering-rag-pipeline is a Claude Code skill for building a local RAG workflow with BM25 retrieval, hierarchical document chunks, and retrieval scoring. It tests retrieval against a prepared set of questions. CX Codex +1 165installs Holographic Memory writerslogic Holographic Memory is an MCP server for Claude Code that stores persistent agent memory locally, with hybrid search, fact forgetting, and optional disk encryption. CX Codex +1 ★ 23repo stars Vector Search — In-Memory TF-IDF Semantic Store axel-belfort Vector Search — In-Memory TF-IDF Semantic Store is an MCP server for Claude Code that indexes and searches small document collections in memory using TF-IDF and cosine similarity. CX Codex +1 3,562installs

Project sponsors

Companies that keep this analytics open