Don't just catch issues.
Resolve them.
When Critique identifies a missing test, an architectural violation, or a logic flaw, it does not just leave a comment. Remedy writes the code, verifies it, and pushes the fix directly to your pull request.
A full developer workflow,
executed autonomously.
An ephemeral VM boots, dedicated to your pull request. Isolated, clean, reproducible.
Downloads the repository, checks the latest git diff, reads the exact structural findings generated by the Critique review.
Writes the required fixes — generates missing test suites, patches security flaws, updates deprecated APIs.
Runs your linting, executes the test suite, confirms the project builds. Nothing broken gets committed.
Commits and pushes working code directly to your PR branch. Critique does one final verification, strictly capped at 2 loops.
Remedy runs on e2b's powerful sandbox.
All in the cloud.
e2b powers our cloud. Our agent downloads your repo, works on it, and submits the PR — entirely in e2b's sandbox. No local setup. No timeouts. Sandboxes stay on for as long as Remedy needs them. If a fix takes 24 hours, the sandbox stays on and our agent keeps working until it's done.
Constrained by design.
Configurable by you.
Already have an AI
coding subscription?
Do not use Remedy at all. Plug in your existing agent instead — Critique acts as orchestrator, generates the execution blueprint, and hands it off. You save every credit for review, not execution.
By using your own agent for execution, you utilise subscriptions you already pay for and reserve Critique credits exclusively for deep multi-model review. If you do not have any of these tools, Remedy handles everything in-platform with no additional setup.
Remedy model floors
These are the model floors only. Final Remedy spend is based on the full autonomous run, so the selected model sets the starting point rather than the final total.
What does one Remedy run actually cost?
A Qwen3 Coder Next run is not just 1 credit. `1 cr` is the model floor. Remedy pricing is based on the full managed run, and if the first pass needs loop 2, one re-review is added on top.
Remedy credits reflect the full autonomous run, not just the model sticker price. The selected model sets the floor, managed execution is always included, and a re-review is added only if the first pass needs loop 2.
Think of the model price as the floor, not the invoice. Final spend depends on how the autonomous run resolves.
Review, fix, and push.
All in one pipeline.
Connect your repo and let Remedy turn review findings into merged fixes.