HOW IT WORKS / THE PRODUCT DIRECTION
One continuous workflow, with evidence at every step and human judgment at the end. This is the experience we’re building.
CAPTURE & PRIORITIZE
The planned starting point is a connected repository and one source of errors: native capture or an existing monitoring provider. Sentry is our first intended integration.
Group related errors and assess their impact. A critical issue and an issue with a safe automated fix are two different things.
INVESTIGATE
Follow the failure into the code, work in an isolated environment, and develop a candidate fix within explicit permissions and spending limits.
The investigating agent must not be able to expand its own authority or change the criteria for publishing its work.
VERIFY & PREPARE
Test the candidate against the failure and gather evidence for review. When the configured policy is met, prepare a pull request that explains the problem, the change, and its verification.
If the result is blocked or inconclusive, report that clearly. A speculative pull request is not progress.
REVIEW & OBSERVE
Your team reviews and merges. ArchLoop is not intended to merge, deploy, or execute code in live production.
The loop closes by observing the outcome after your deployment. An opened pull request, a merged change, and a resolved production issue are distinct results.
WHERE WE ARE TODAY
Sign in with GitHub to open ArchLoop. We’re still validating the complete error-to-pull-request workflow. Automated investigations and paid plans are not enabled yet.
Our initial focus is Node.js applications using GitHub. Browser, Next.js, and source-map support remain planned; broader framework compatibility will be validated before we offer it.
Open dashboard