BugRep-AI
IBM Bob…
Autonomous bug repair · powered by IBM Bob

BugRep-AI
From bug report to verified repair.

Describe a bug in plain English. AI agents find the defect, prove it with failing tests, write a fix and prove it with the same locked tests. Nothing touches your code until you approve.

Analyze my code
Isolated sandbox copy Tests locked by SHA-256 Human approval required

Analyze & repair

Choose code, describe the bug, pick an agent.

1

The folder open in your IDE. BugRep works on a sandbox copy; nothing changes until you approve.

Public repos work out of the box. For private repos set GITHUB_TOKEN in .env.

2
Expected behaviour / business rules (optional)
3
… Starting
··0s
New run
    Bug report

    Activity

    Runs

    Every run ends in an explicit state. Nothing retries automatically.

    AI agents

    IBM Bob is the operational agent. Other providers are shown for the roadmap and are never called.

    How multi-agent review will work

    InvestigatorsA · B · C investigate independently
    Compare findingsagreement on file / function
    Repairerseach proposes a candidate
    Test runner decidessame locked tests; no AI vote

    The orchestration interface is in place (agents/index.js). It turns on when two or more agents are operational.

    Integrations

    Optional. Configured on the server in .env; secrets never reach the browser. Delivery never changes a repair's status.

    Reports

    Evidence reports (Markdown + JSON) for every finished run.