ai-memory MCP
ai-memory retains session knowledge in a Markdown wiki, exposes MCP retrieval and passes context between agents and machines.
- 15,226 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Hooks retain work-session information and the shared server exposes project knowledge to connected users and agents.
Install
- Works in
Manual install
curl -fsSL https://github.com/akitaonrails/ai-memory/releases/latest/download/install.sh | sh
ai-memory setupFirst Linux/macOS Quick start block. Inspect the script and setup plan before agreeing on integrations.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Hooks retain work-session information and the shared server exposes project knowledge to connected users and agents.
This is third-party code. Review the repository files before installing.
What it does
When switching agents, the next session receives a note describing where the task stopped, what failed and which questions remain open. ai-memory records connected-session events in a project Markdown wiki. Git preserves note versions, and the search index can be rebuilt from those files. The server runs locally or in selected infrastructure. Teammates search shared project knowledge, while notes about unfinished personal tasks belong to individual users. Event recording, search and task continuation work without a model key; model-based note processing is optional.
Who it is for. Teams switching between Claude Code, Codex and other agents while retaining inspectable project memory.
Good fit when
- A Claude Code session ends and the user continues the task in Codex.
- Work moves from a desktop to a laptop without the previous context.
- Teammates share project decisions while keeping unfinished-task notes personal.
- The owner corrects a mistaken decision in an ordinary Markdown editor.
Not a fit when
- Work-session events and prompts may not be retained.
- The server would be exposed without authentication review.
- Compacted or merged notes must be treated as original evidence.
Example request
Show an ai-memory plan for one test project, exclude secret paths and connect only the agent selected by the owner.MCP
- Transport
- stdio, http
- Authentication
- API key
Limitations
Setup connects a service, hooks and MCP to selected agents, so inspect its plan first. Capture sanitizes observations, but configure project and path exclusions for your own data. Older installations and the Docker quick start may lack authentication; inspect status before granting remote clients access. Optional model transformations and vector search need separate configuration, including review of which notes reach providers and how their outputs are validated.
How to disable. Run `ai-memory uninstall --apply` to remove components installed by ai-memory and clear its agent auto-wiring record. A later `ai-memory run` will wire the agent again; use `--no-autowire` or `AI_MEMORY_RUN_AUTOWIRE=false` to prevent that.
FAQ
How do I find recent notes?
Run `ai-memory recent` for the current project. Add `--global` for recent pages across projects, then ask the agent about the decision or open task you need.
What does another agent receive when continuing a task?
A brief of the latest work state, failed attempts and open questions. The note has an owner and can be claimed once; further details are retrieved from shared project memory.
Can I preview setup changes?
Yes. `ai-memory setup --plan` lists detected agents, directories and planned changes without writing them. Select the intended integrations before applying setup.
Why inspect older installs?
An older installation may have token authentication disabled. Check `ai-memory status` before granting remote clients access to the server.
What is available in Claude Desktop?
Memory reads and writes through MCP. This integration has no automatic session hooks, so Claude Desktop events are not recorded in the same way as Claude Code events.
Discussion
No questions or reviews yet.
Related
MCP serversiai-memory
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
Long-term memory for AI agents: Markdown files as the source of truth, a SQLite index as a cache, shared by Claude Code, Codex and others
A long-term memory MCP server for AI agents: facts fade along a forgetting curve and related records are compressed into a summary
A long-term memory system for AI agents: the agent stores experience and facts, then retrieves them through a built-in MCP server
Local memory for coding agents: sessions turn into a git-versioned Markdown wiki and come back through an MCP server.
Serein keeps an AI's long-term memory on your own server and serves it to chats through MCP tools and an OpenAI-compatible gateway
README in short
Built-in user management separates shared project knowledge from personal tasks. Writes retain authorship, and an audit log records memory operations.