Skip to content

How it works

From issue to merged PR, without the babysitting.

developerz.ai turns scoped GitHub issues into reviewable pull requests through an audit-first loop your team can inspect at every step.

Join the waitlistRead the quickstart
acme/api · Issue #482LIVE
Handle expired refresh tokens

Return a typed error, add regression coverage, and update the authentication guide.

developerzbugauth

developerz-bot I've mapped this to session.ts and two integration tests. Estimated cost: $1.42.

Audited end to endGitHub-native workflowExisting checks stay in controlCost visible per task

The run

A predictable path from issue to merge.

Six stages, each receipted. Your team can inspect every action or stop the run at any point.

  1. 01
    Issue assigned

    Label an issue or mention the dev. The issue stays the source of truth.

  2. 02
    Context mapped

    It reads the repository, the thread history, and your .maintainer.yml before acting.

  3. 03
    Plan approved

    The approach posts to the issue first, so a human can stop it before code changes.

  4. 04
    Changes authored

    Coding hands off to your own agent (Copilot, Claude Code, or OpenHands) on your keys.

  5. 05
    CI verified

    Branch protections stay intact; every required check must pass before merge.

  6. 06
    Reviewed and merged

    Machine gates merge green PRs; every step lands in the hash-chained audit log.

Guardrails

Boundaries are part of the brief.

Your .maintainer.yml defines what the dev may touch: versioned, reviewed, and enforced on every run.

  • Allowed repositories
  • Protected file paths
  • Required CI checks
  • Human approval gates

Questions

Frequently asked questions

Does the bot pretend to be human?

Never. Every comment it posts discloses that it's an automated agent. Non-disclosure is a bug, not a setting.

Whose API keys does it use?

Yours. BYOK is the only mode, your own Anthropic, OpenAI, Gemini, Kimi, GLM, MiniMax, DeepSeek, xAI, Groq or OpenRouter key, or any OpenAI-compatible endpoint. We never resell tokens.

Does it write the code?

No. It's a thin orchestrator: it triages and coordinates, then hands coding to your agent (Copilot, Claude Code, or OpenHands).

Can I see what it did?

Yes. Every tool call and decision is logged. If it isn't logged, it didn't happen.

What about CodeRabbit or Dependabot?

It detects other bots and defers: no bot-on-bot loops. It waits for their reviews to settle before acting.

Put a fleet of AI developers to work

Bring your own keys and your own boxes, drop a .maintainer.yml, and let the loops run. Audited end to end. Access is invite-only.

Join the waitlistRead the quickstart