Quickstart

This is the fastest path to a running Memory Layer with real memories to explore. It uses the bundled Docker stack, so you don't need to install PostgreSQL, pgvector, or configure anything by hand. It works keyless — no API keys required.

1. Start the stack

From a clone of the repository:

docker compose up

That builds and starts three things together: PostgreSQL with pgvector, the Memory Layer service (which applies its database migrations on boot), and the web UI. When it settles, the service is listening on http://localhost:4040.

Open http://localhost:4040 to see the web UI. It starts empty — the next step fills it.

2. Load the demo project

In another terminal:

memory demo

This seeds a project called demo with a relatable set of memories about Memory Layer itself — its architecture, decisions, conventions, and more — then curates them. Now every surface has something real to show.

docker compose exec memory memory demo

3. Ask it something

memory query --project demo --question "How does reinforcement work?"
memory query --project demo --question "What does the service need to run?"

You'll get a short, cited answer drawn from the demo memories — with a confidence score and an honest signal when the evidence is thin.

4. Explore

memory tui

The terminal UI lets you browse and search every memory, inspect where each came from, and open the memory graph. The web UI at http://localhost:4040 shows the same data with wider panes and a 3D graph view.

What you only need to know

Three commands cover day-to-day use:

• memory remember — capture and curate completed work into memory.
• memory query — ask a project-specific question against curated memory.
• memory resume — get a briefing to pick up where you left off.

Everything else is grouped under memory --help, and the full install guide covers native (non-Docker) setups, LLM and embedding providers, and connecting your coding agent.

Next steps

• Three things to know — the whole mental model on one page.
• How it works — memories, retrieval, reinforcement, and the code graph.
• Memory types — how knowledge is classified.
• Install guide — native install, providers, and agent setup.
• Automations — the built-in loop workflows.