OMEGA Memory MCP
Local long-term memory for agents: decisions, mistakes and preferences are stored on your machine and pulled into new sessions
- 225 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- OMEGA hooks are written into `~/.claude/settings.json` and save session context to disk where secrets from the conversation may end up
- The consult tools send context to another model provider with your key
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add mcp/omega-memoryDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- OMEGA hooks are written into `~/.claude/settings.json` and save session context to disk where secrets from the conversation may end up
- The consult tools send context to another model provider with your key
This is third-party code. Review the repository files before installing.
What it does
Coding agents do not remember past sessions: decisions, agreements and mistakes have to be explained again. OMEGA keeps them as memory that runs entirely on your computer, including computing embeddings (ONNX), with no cloud and no API keys. It works with Claude Code, Cursor, Codex, Windsurf, Claude Desktop and any MCP client and is not tied to one model provider. By default the client sees five tools so as not to burn context: `omega_store`, `omega_welcome`, `omega_protocol`, `omega_tools` and `omega_call`, through which the others are opened. The main set: `omega_store` saves a decision, lesson, error or preference, `omega_query` searches by meaning, exact phrase or time, `omega_welcome` gives a session briefing, `omega_checkpoint` and `omega_resume_task` pass task state between sessions, `omega_reflect` finds contradictions, `omega_review` checks a diff with the project's conventions in mind and `omega_remind` sets reminders. Claude Code hooks capture context automatically.
Who it is for. Developers who work on one project in Claude Code, Cursor or Codex for weeks and are tired of repeating architectural decisions and preferences to the agent.
Good fit when
- The agent again proposes a scheme you rejected a week ago and OMEGA recalls the reasons for rejection
- A session broke off in the middle of a task and `omega_checkpoint` saves its state while `omega_resume_task` continues it
- Code style and naming preferences are recorded once and pulled into all projects
- Switching the model from Claude to GPT should not wipe the accumulated memory
Not a fit when
- You need memory shared by a team on a server: OMEGA targets a single machine and team tools are in the paid Pro version
- You need an instant solution with no install: setup downloads models and installs hooks into `~/.claude/settings.json`
- Python is older than 3.11: the package needs 3.11 or newer
Example request
Recall why we rejected the message queue in this project and save the new decision: we use a simple task table in the databaseMCP
- Transport
- stdio
- Authentication
- not required
Limitations
Python 3.11 or newer is required (macOS ships 3.9, so install a fresh one first). Installation goes into a separate environment: `pipx install "omega-memory[server]"`, the quotes are mandatory in zsh. Without the `[server]` extra there is no MCP server and no hooks. The `omega setup` command downloads the bge-small-en-v1.5 embedding model (about 130 MB) and a reranker (about 90 MB) into `~/.cache/omega/models/`, registers MCP and installs hooks into `~/.claude/settings.json`. `--dry-run` previews it. Coordination, routing and a knowledge base are in OMEGA Pro. The license is Apache-2.0.
How to disable. Run `claude mcp remove omega-memory` and remove the OMEGA hooks from `~/.claude/settings.json`, then uninstall the package (`pipx uninstall omega-memory`). Data lives in `~/.omega/` (or `$OMEGA_HOME`) and is deleted by hand.
FAQ
How do I install OMEGA?
With `pipx install "omega-memory[server]"` (or `uv tool install "omega-memory[server]"`), then `omega setup` and `omega doctor`. Python 3.11 or newer is required.
How do I connect it to Claude Desktop?
Run `omega setup --client claude-desktop`: it registers the server in the Claude Desktop config, after which a restart is needed. For Codex and Antigravity setup writes the config file and for other clients it prints a JSON block.
Does OMEGA send data to the cloud?
No, storage and embeddings run on your machine and no keys are needed. The exception is the `omega_consult_gpt` and `omega_consult_claude` tools, which send context to another provider with your key.
Why does the client see only five tools?
That saves context. The others are discovered through `omega_tools` and called through `omega_call`, and storage, search, checkpoints, reminders and maintenance are available.
What does the omega doctor command do?
It checks every installation step and names the fix for anything that failed. For hooks there is a separate `omega hooks doctor`.
Discussion
No questions or reviews yet.
Related
MCP serversLocal memory for Claude Code, Codex and Claude Desktop: one Markdown folder and one MCP server shared by all.
An MCP server that gives an agent memory across sessions, projects and machines: decisions, dead ends, preferences and skills
Persistent agent memory in a folder of linked markdown notes, so Claude Code remembers a project across sessions
Local persistent memory for coding agents: hooks record every session, and Claude Code, Codex or Cursor recall past decisions through 54 MCP tools
A long-term memory MCP server for agents: one Rust binary that runs offline and never calls a language model
Shared memory for Claude Code, Codex and Cursor: a decision made in one session is recalled in another tool or on another machine.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| OMEGA Memory (this card) | no | not required | 225 | |
| MCP Memory Service | no | not required | 2,117 | a self-hosted memory with a knowledge graph and REST API rather than local ONNX memory with Claude Code hooks |
| agentmemory | yes | not required | 30,340 | hooks record the course of every session while OMEGA stores decisions, lessons and preferences with task checkpoints |
README in short
The text lists the problems of agents without memory: context loss, repeated mistakes, someone else's cloud instead of your own database and vendor lock-in. Next come installation, `omega setup` with an `omega doctor` check, client connections, the SQLite storage design, a reference of tools and CLI commands and a comparison with cloud memory.