Advisor vs. Supervisor
Rigour operates in two distinct modes depending on how you integrate it into your workflow.
1. Advisor Mode (MCP)
In Advisor Mode, Rigour acts as a high-fidelity consultant for your AI agent.
- Role: Pre-flight Validator.
- Trigger: The agent proactively calls Rigour tools (e.g.,
rigour_check) to verify its work. - UX: Seamlessly integrated into IDEs like Cursor or Cline.
- Philosophy: Agent-driven. The agent asks: "Is this code good enough to commit?"
When to use:
- Interactive development in an IDE.
- Complex refactors where the agent needs frequent feedback before finishing a turn.
2. Supervisor Mode (CLI Loop)
In Supervisor Mode, Rigour acts as the "governor" of the entire process.
- Role: Iterative Controller.
- Trigger: You run
rigour run -- <agent-command>. - UX: Terminal-based, wrapping agents like Claude Code.
- Philosophy: Tool-driven. Rigour says: "You are not finished until these gates pass. Here is what you failed."
When to use:
- Headless automation or CI/CD pipelines.
- Terminal-based coding agents.
- Ensuring 100% compliance without manual re-prompting.
Comparison Summary
| Feature | Advisor (MCP) | Supervisor (CLI) |
|---|---|---|
| Primary Tool | @rigour-labs/mcp | @rigour-labs/cli |
| Control Flow | Agent calls Rigour | Rigour wraps Agent |
| UX | VS Code / JetBrains / Desktop | Terminal / Shell |
| Feedback | Interactive Response | Loop Iterations |
| Example Agent | Cursor, Cline, Gemini Desktop | Claude Code, Shell Agents |