Pilot

One question, several models, and the places they disagree.

Pilot asks several independent AI models the same question, lets them critique each other, and gives you a decision with the dissent attached instead of one model's confident guess.

Open Pilot View on GitHub

What it does

Models answer independently, then see and rebut each other's positions across rounds.
A synthesizer reports a final answer, a consensus level (full, majority or split) and any named dissent.
Goes beyond Q&A into human-gated software work: discover, ideate, approve, draft, negotiate, execute, review, publish, outcome, with every step in an append-only ledger.
Works inside your editor as an MCP server for Claude Code, GitHub Copilot and Cursor.