iai-memory MCP
iai-memory (Personal Memory Engine)
A local long-term memory MCP server for agents: keeps conversations verbatim and returns the right context at the start of a session
- 923 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Runs a background service and stores conversations on disk
- Captures session content verbatim, including sensitive data
Install
- Works in
Manual install
python3.12 -m pip install -U iai-pmeThen in Claude Code run /plugin marketplace add CodeAbra/iai-personal-memory-engine and /plugin install iai-memory@iai-pme.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Runs a background service and stores conversations on disk
- Captures session content verbatim, including sensitive data
This is third-party code. Review the repository files before installing.
What it does
iai-memory gives an agent long-term memory through a local MCP server. It captures every turn of every session verbatim, over time builds those captures into a map of who you are and how you work, and returns a small slice of relevant memory at the start of a new conversation. Data is stored locally and can be encrypted, framed as an alternative to external RAG. The project is packaged as a plugin with an MCP server and a background service, with a Rust engine and Python and TypeScript wrappers. It works with Claude Code, Cursor, Codex and other MCP clients, installing via PyPI and the plugin marketplace or a single bootstrap script.
Who it is for. For those who want an agent to remember past conversations and use them without an external store.
Good fit when
- You need long-term agent memory across sessions
- You want context stored locally, not in the cloud
- You want an alternative to external RAG for personal context
Not a fit when
- You do not need memory across sessions
- You cannot run a background service or build a native engine
Example request
Recall what we decided about this project's architecture last timeMCP
- Transport
- stdio
- Authentication
- not required
Limitations
It needs Python 3.11 or 3.12, and building from source requires Rust and extra tools. The server runs as a background service and stores data locally, with no cross-machine sync. Verbatim capture means sensitive data also lands in memory, so encryption and access control matter.
How to disable. Uninstall the iai-memory plugin via /plugin, stop the background service and remove the iai-pme package.
FAQ
Where is data stored?
Locally on your machine, and memory can be encrypted, framed as an alternative to external RAG.
What exactly is remembered?
Every turn of a session verbatim, and over time the captures form a map of how you work.
Discussion
No questions or reviews yet.
Related
MCP serversai-memory keeps long-term assistant memory in local SQLite and serves it over MCP, an HTTP API and a CLI
Long-term memory for agents: a remote MCP server, plugins for Claude Code, Cursor and Codex, and an API for your own apps
A long-term memory system for AI agents: the agent stores experience and facts, then retrieves them through a built-in MCP server
An MCP server for long-term agent memory: compresses dialogues into atomic facts, resolves pronouns and anchors relative time to absolute dates
A self-hosted MCP server for long-term agent memory: structured records, versioned rollback and a visual dashboard
A local-first MCP memory server for agents: session history, codebase index and a concept graph in one SQLite database
README in short
The README describes iai-memory as a local MCP server that gives an agent long-term memory: it saves conversations verbatim, builds a user map from them and returns a relevant slice at the start of a session. Storage is local with optional encryption, and the engine is Rust with Python and TypeScript wrappers. Install is via PyPI and the plugin marketplace for Claude Code or a single bootstrap script for macOS and Linux. It works with Cursor, Codex and other MCP clients. MIT licensed.