# Claude Code

Give Claude Code a memory of your project: decisions, people, past fixes.

## Connect

```
claude mcp add --transport http remiq https://<your Remiq address>/mcp
```

Use your MCP address from the Console → **Agents**. Then type `/mcp` in Claude Code and sign in: a Remiq page asks for an API key and the memory space to use.

## Use it

- Before a task, the agent can read `prism_context` with the task: past decisions, the people involved and what was tried.
- After a fix, it can save what worked with `prism_add`.
- For a repository Remiq has indexed, `prism_map` shows what a change affects.

Building your own coding agent? Log each run with [Agent runs](/docs/agent-runs/): a later context cites the past fix for the same code.

## Add memory to the app you are building

Copy the [setup prompt](/setup-prompt.md) into Claude Code. It tells Claude Code how to wire Remiq into your app: keys from the environment, a save per message, a context call before every model turn.