Get a second opinion from other AI models

Ask other AI models to critique your plan, diff or pull request — preview the cost first:

agent-config council:estimate "Should the export run in a background job?"

In your AI tool:

/council:default prompt:"Should the export run in a background job?"
/council:pr 482

What it does

Agent Config sends your question — a prompt, a roadmap, a diff or a set of files — to models from other providers than the one you are working with, then collects their answers side by side. Each member sees a neutral, redacted framing, not your agent’s opinion of the answer. The result is advice; nothing is changed or posted for you.

/council:default — a plan, a prompt, a roadmap, a diff or files.
/council:pr <number> — a GitHub pull request; read-only unless you ask it to comment.
/council:design, /council:optimize, /council:debate — the same with a lens for design, optimisation or a structured debate.
/roadmap:ai-council — challenges a roadmap before you start it.

When to use it

For decisions that are expensive to get wrong: an architecture choice, a migration plan, a pull request on a sensitive path. Not for routine changes — every call costs money.

What it needs

An API key for Anthropic, OpenAI or both, stored once for your user:
Terminal window
```
agent-config keys:install-anthropic
agent-config keys:install-openai
```
agent-config council:status shows whether a council is configured, without spending anything.

Cost

council:estimate previews the cost and makes no API call. council:run spends only with --confirm, and a daily per-provider cap stops runaway use. Your keys and the council settings live in your user folder, ~/.event4u/agent-config/, never in a repository.

Limits

Other models disagree too: a split answer is information, not a verdict.
Your question, and the files or diff you choose, are sent to the providers you configured. What leaves your machine.