Ami is unclaimed —
A
Ami
Ami fixes the hardest bugs your coding agent can't.·ami.dev
Ami is a desktop application that runs a team of AI agents to find and fix bugs and improve code quality. It reproduces bugs by instrumenting and running code locally, spawns parallel agents to test competing root-cause hypotheses, then patches and validates the fix. It also reviews pull requests by checking out and running the branch to give feedback based on actual runtime behavior.
What it's for
Fixing bugs that a coding agent could not resolveReproducing bugs with real runtime data instead of static code analysisDiagnosing root causes via parallel agent hypothesis testingReviewing pull requests by running the branch and reporting on actual behaviorPre-checking your own PRs before pushing to catch issues reviewers would flagCatching React infinite re-render loops, SSR hydration mismatches, race conditions in streaming responses, and performance bottlenecks
Features 14
- Accepts a bug report or Sentry alert as input
- Catches race conditions in streaming responses
- Catches React infinite re-render loops
- Catches SSR hydration mismatches
- Checks out and runs a PR branch to give review feedback based on actual behavior
- CLI installation via curl script
- Detects performance bottlenecks and unnecessary re-renders
- Instruments and runs code locally to reproduce actual bugs (no static-code guessing)
- Login/setup flow on first run
- Patches code and validates that the fix resolves the issue
- Pre-checks your own PRs before pushing
- Runs entirely on the user's own machine
- Spawns parallel agents to test competing root-cause hypotheses
- Supports frontier models including Claude Opus, Claude Haiku, GPT-5.2, Gemini 3 Pro, GLM, and Kimi K2.5
Pricing
Pricing is usage-based, tied directly to underlying model API costs with no markup. First 2M tokens are free. Rates vary per model.
Free to startFree
2M tokens free to start
Pay as you go (API cost)Quote
We don't charge over API cost