gauntlet

Run your codebase through the gauntlet: 55 specialized review prompts, dispatched to whichever AI coding agents you have installed, applying fixes directly to the working tree.

Download GitHub

Screenshot of the gauntlet terminal dashboard: an activity chart across the top, one row per agent lane with its live token rate, a grid of the scheduled reviews each with a status glyph, and a normalized output feed

What it does

One static binary loops 55 review prompts over your repository, hands each one to an agent CLI you already have installed, and applies what it finds.

Isolation in parallel. --jobs N gives N lane worktrees pulling from a shared queue, each review on its own branch, merged back one at a time.
Unmerged review stacks. --stacked-prs opens each changed review as a child PR of the previous one. The branch in your editor never changes.
A dashboard. --tui puts lanes, throughput, the review grid, and a normalized feed on one screen.
Measured, never guessed. Token counts come from what the agent reports; an agent that reports nothing shows no rate.
A record. Every run is JSONL under ~/.gauntlet, replayable with gauntlet show.

Quick start

gauntlet doctor     # which installed agents it can use
gauntlet pick       # compose a run
gauntlet -r quick --stacked-prs --pr-base main

Works with claude, codex, gemini, qwen, cursor-agent, opencode, and more. Linux and macOS. Install steps are in the README.