Inspiration
I maintain open-source projects. Bounties are a good way to get help, but paying them is slow manual work. You read every merged PR, decide whether it really fixed the issue, look up the contributor's payment details, send the money, and keep a record. It's slow and error-prone, and contributors wait weeks.
I wanted that settlement step to run on its own without handing my judgement, or my money, to a black box.
What it does
A maintainer labels an issue bounty:50. A contributor opens a PR that says Fixes #12 and PayPal: [email protected]. When the PR is merged:
- Claude reads the issue, the PR and the diff. It returns a structured verdict (pay or hold), a confidence score, and reasoning a maintainer can check.
- Code applies the mandate. If Claude is at least 80% confident and the bounty is $100 or less, BountyPay pays the contributor through PayPal Payouts and comments a receipt on the PR explaining why.
- Everything else is held. The maintainer sees it in an AG Grid dashboard with Claude's full reasoning in a side panel, and approves or rejects after a confirmation.
- PayPal reports the outcome. Payout status comes from PayPal webhooks, with a status lookup as a fallback. "Paid out" counts only money PayPal confirms.
In the live demo:
- Claude auto-paid a real fix at 92%.
- A typo-only "fix" was held at 2%.
- A $250 bounty was held because it's over the cap, whatever Claude's confidence.
- A PR that tried to talk Claude into approving itself was held at 2%. Claude's reasoning names the manipulation.
A bounty is paid at most once, however often a webhook is redelivered or a button is double-clicked.
Built With
- claude
- next.js
- offline-mode
- paypal
- postman
- render
- typescript
- webhooks
