Metronix Memory MCP
Self-hosted agent memory: an MCP server stores facts and preferences and searches them with a Neo4j graph and citations.
- 107 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- Memory keeps facts and preferences in a database on your server, and the installer runs through `curl | bash`.
Install
- Works in
Manual install
curl -fsSL https://mtrnix.com/install.sh | bashThe README one-liner for the latest release. Fallback: clone the repository and run `./install.sh`.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Memory keeps facts and preferences in a database on your server, and the installer runs through `curl | bash`.
This is third-party code. Review the repository files before installing.
What it does
Metronix Memory is memory infrastructure for AI agents that deploys with Docker on your side. An agent saves facts, preferences and pinned context scoped to a workspace and agent, then finds them. Search combines dense and sparse vectors (`SPLADE`) with `Neo4j` graph context, and answers carry source citations. The MCP server is available at `/mcp` with the tools `metronix_memory_store`, `metronix_memory_search`, `metronix_memory_list`, `metronix_status` and `metronix_search_fast`. Besides MCP there are REST and an OpenAI-compatible API. Files and SaaS data load into the knowledge base and refresh as projects change.
Who it is for. Teams running several agents who want shared long-term memory on their own server rather than a vendor cloud.
Good fit when
- An agent must remember user preferences across sessions.
- Shared project knowledge is needed by all agents with a source named.
- Data cannot go to an external memory service.
- Hermes and other runtimes attach to one store by agent identifier.
Not a fit when
- The machine has under 6 GB of memory: the stack needs at least 6 GB, ideally 8, and about 15 GB of disk.
- A `CLAUDE.md` file or a simple note is enough.
- Running Docker and a Neo4j graph for one agent memory is too much.
Example request
Remember that the staging database password rotates quarterly and find that note with memory search.MCP
- Transport
- http
- Authentication
- API key
Environment variables
METRONIX_MCP_API_KEYrequired, secret- Shared key for /mcp from the server .env file
AGENT_UUIDrequired- Agent identifier for the X-Agent-Id header
Limitations
The stack needs Docker with at least 6 GB of memory (8 recommended) and about 15 GB of disk, and the first image build takes 10-15 minutes. The shared `METRONIX_MCP_API_KEY` opens `/mcp` only, while memory tools and REST need a personal `mtk_` key. The `agent_id` argument must match the `X-Agent-Id` header. Install is described as a `curl ... | bash` line that downloads a script from the network: read the script before running. License Apache-2.0.
How to disable. Stop the Docker Compose stack and remove the `/mcp` entry from the agent config. Delete data volumes and keys separately.
FAQ
How do I connect an agent?
Two paths: paste prepared prompts into the agent or register the MCP connection by hand. Both are described in `connecting_to_agent.md`.
What format does `agent_id` have?
1 to 64 characters from letters, digits, dot, underscore and hyphen. Spaces and slashes are rejected.
Is a separate key needed for memory tools?
Yes, a personal `mtk_` key. The shared `METRONIX_MCP_API_KEY` works for `metronix_status` and `metronix_search_fast`.
What is the default workspace?
The `DEFAULT_WORKSPACE_ID` setting, `MTRNIX` by default. Workspaces are listed in the UI or via `GET /api/v1/workspaces`.
Is there an answers mode?
Yes, installing with `--mode answers --openwebui -y` adds a chat UI and answer generation.
Discussion
No questions or reviews yet.
Related
MCP serversA self-hosted MCP memory server that keeps preferences and decisions across conversations in Claude Code, Cursor and other clients
Self-hosted long-term memory for agents: semantic search, a knowledge graph and autonomous consolidation over MCP and a REST API
An MCP server for long-term agent memory: compresses dialogues into atomic facts, resolves pronouns and anchors relative time to absolute dates
A long-term memory MCP server for agents: search, store and correct facts across sessions in the cloud or a local core
A self-hosted MCP server for long-term agent memory: structured records, versioned rollback and a visual dashboard
A long-term memory MCP server for AI agents: facts fade along a forgetting curve and related records are compressed into a summary
README in short
The README gives a quick start, a five-minute store-and-find memory example, a comparison with vector databases and links to install, agent connection and MCP reference.