Claude Code

Connect Claude Code so every task it works is constrained by your architectural decisions. You need Trurlic installed with .trurlic/ initialized, and Claude Code (npm install -g @anthropic-ai/claude-code).

Setup

trurlic install --ide claude-code

This runs claude mcp add trurlic -s user -- trurlic serve, registering the MCP server at user scope.

Share it with your team via git — register at project scope instead, which writes .mcp.json at the repo root so every clone gets it:

claude mcp add trurlic --scope project -- trurlic serve

If trurlic isn’t on PATH, pass an absolute path (which trurlic) as the last argument.

Agent instructions

Add to your project’s CLAUDE.md:

## Trurlic

This project uses Trurlic for architectural decisions.

Before implementing any task:
1. Call `advance` with the component name and a mode (agent | interactive).
2. Follow the returned action exactly, calling `advance` again after each.
3. Repeat until `ready: true`, then implement within the context brief.
4. Call `verify_against_decisions` before committing; fix any violation.

For the full recommended instructions — including agent vs interactive mode — see the trurlic repo’s CLAUDE.md.

Verify

1. trurlic init && trurlic add component auth -d "Authentication"
2. trurlic install --ide claude-code, then start a new Claude Code session.
3. Ask it to design the auth component — it should call advance → get_step_prompt and walk you through the Socratic flow.
4. Ask it to implement auth — it should call get_context and follow the recorded decisions.
5. trurlic status shows the decisions it recorded.