Skip to content
MMW Docs

Quickstart

Copy page Markdown llms.txt

By the end of this page you will:

  1. have an MMW API key;
  2. have MMW connected to your AI assistant;
  3. save a fact in one session and find it in another, without reminding the assistant yourself.

The example used throughout: your team agreed that staging is deployed with the systemd unit app-staging. A week later a new developer asks the agent how to deploy staging, and the agent answers correctly because it remembers the agreement.

1. Account and key

Section titled “1. Account and key”

  1. Sign up at app.mmwhub.tech and confirm your email with the link we send you.

  2. Choose a plan. Free is a 3-day trial, no card required.

  3. Copy your API key. It starts with mmw_ and is shown only once, so store it in a password manager.

2. Connect your assistant

Section titled “2. Connect your assistant”

The recommended way is the local agent mmw-agent: it exposes every MMW tool and, if you allow it, uploads your session history. You need Python 3.11+ and uv (the uvx command).

Terminal window
claude mcp add mmw \
  -e MMW_API_KEY=mmw_your_key \
  -e MMW_ENDPOINT=https://mcp.mmwhub.tech \
  -- uvx --from https://app.mmwhub.tech/downloads/mmw-agent/mmw_agent-0.1.4-py3-none-any.whl mmw-agent

Restart the client. The tool list should now include remember, search, forget, validate_memory and gateway_call, plus mmw_sync_status when you use mmw-agent.

3. Save a fact

Section titled “3. Save a fact”

Ask your assistant:

Remember in MMW: staging is deployed with the systemd unit app-staging. Fact key: deploy-staging.

The assistant calls remember. The server response looks like this:

{
  "id": "4d428a41-e7b0-4f81-a886-f40c6f6c766c",
  "workspace": "default",
  "scope": "shared",
  "status": "unverified",
  "source_id": null,
  "derived_from": [],
  "created_at": "2026-10-05T09:12:00+00:00"
}

unverified is the normal status of a new record: MMW keeps track of where a fact came from, but does not claim it is true. Read more in Source and freshness.

4. Find it in a new session

Section titled “4. Find it in a new session”

Close the conversation, open a new one (it can be a different client with the same key) and ask:

How do we deploy staging? Check MMW.

The assistant calls search and gets the record together with its source and status:

{
  "result": [
    {
      "id": "4d428a41-e7b0-4f81-a886-f40c6f6c766c",
      "content": "staging is deployed with the systemd unit app-staging",
      "fact_key": "deploy-staging",
      "status": "unverified",
      "confidence": 0.5,
      "graph_relations": []
    }
  ]
}

The record also appears in your account under Knowledge base.

How it fits together

Section titled “How it fits together”
  Session 1               MMW                        Session 2 (any client)
  ─────────               ───                        ──────────────────────
  remember(fact)   →   record + source       →     search(question)
                       + status + links             → fact with source and status
                       (isolated per account
                        and key project)

Memory is not tied to a model or a computer: Claude, Cursor and Codex using the same key see the same records.

What’s next

Section titled “What’s next”
How MMW works Records, projects, statuses and the archivist: the big picture.
What to remember How to get the agent to save useful facts instead of noise.
All clients Detailed setup for each assistant.
Session history Upload Claude Code and Codex sessions, only with your consent.