What is Pullfrog?
Pullfrog is an open-source, agent-agnostic harness for asynchronous development inside GitHub. It provides a standard interaction model for any team to automate planning, coding, and review using their preferred agent. Once installed, you can trigger agent runs by tagging@pullfrog anywhere in your repo, from the web dashboard, or via automated triggers.
Key features
- 🔌 Bring your own model — Powered by a secure, sandboxed agent harness that works with any LLM provider: Anthropic, OpenAI, Google, Mistral, DeepSeek, OpenRouter, and more.
- 💬 Tag your agent anywhere — No new UI to learn. Trigger agent runs with an @mention anywhere in your repo: issues, PRs, comments.
- ⚡ Automatic triggers — Configure automated responses to events like new issues, new PRs, CI failures, or PR reviews.
- 🧠 Intelligent mode selection — Pullfrog automatically selects the right mode based on context: Build for implementation, Review for code reviews, Plan for task breakdowns, or Address Reviews for responding to feedback.
- 🔄 Self-healing PRs — Pullfrog automatically addresses CI failures and PR reviews on its own PRs.
- 👥 Built for teams — Organization-wide permissions, shared configuration, and async workflows. GitHub remains the single source of truth for access control.
- 🔐 Secured by GitHub — All API keys are stored as GitHub secrets. Actions are performed using short-lived tokens that auto-revoke when the run completes.
- ⚙️ Powered by GitHub Actions — Runs entirely in your GitHub Actions using your own API keys. You control the infrastructure, the secrets, and the costs.
- 🆓 Free for open source — The Pullfrog GitHub action is open-source and free for any open source projects, forever.