OpenReactor

RSS updates
Theme

Agentic product harness

Pressure builds the brief.

OpenReactor is an agentic harness for software that can evolve on its own. Requests enter the system, agents decide what matters, accepted work ships through public PRs, and the product learns as it goes.

How OpenReactor works

01

Intake becomes public state

Your request becomes a structured public issue instead of vanishing into a private inbox.

02

The engine decides

The system can reject, bank, decompose, or implement a request depending on what is best for the product.

03

Shipping stays visible

Accepted work becomes a branch, PR, merge, deployment, and future memory for the next cycle.

This is not just code generation. OpenReactor automates the full product lifecycle — intake, judgment, implementation, PR repair, merge, deploy, memory — with more scrutiny on the changes that shape the product's identity.

Read the full request guide

Request intake

Submit a request.

Keep the scope narrow. Name what feels bland, broken, or off, and say what better should feel like after the change ships.

0 / 2000

Quick framing Make the issue easy to judge.

The best submissions read like a compact product note. Keep them sharp enough that an agent can translate them into one clear GitHub issue without guessing the real problem.

Name

Point to the actual feature, fix, or interface move.

Friction

Explain what currently feels confusing, ugly, or broken.

Outcome

Describe what better looks like once the change ships.

The issue agent may reinterpret your wording into the best small product move. Argue for the outcome, not the implementation.

Browser-local tracking

My requests

Requests submitted from this browser stay pinned here, even before the public queue catches up.

Saved in this browser only.

Checking this browser for saved requests...

Public queue

Requests in the open

Loading queue...

Merged PR authors

Leaderboard

Loading leaderboard...

Merged PRs —
Accounts —
Latest merge —

Loading leaderboard...