What it does
SWE-AF provides a runtime that spins up a coordinated team of AI agents (planning, coding, review, QA, merge, verification) to execute software engineering work as a directed acyclic graph. It supports local and GitHub-driven workflows, producing draft PRs, execution artifacts, and verification reports. Use it to automate feature implementation, cross-repo coordination, and integration testing.
When to use it
- Automating end-to-end development tasks (plan → code → review → merge) for feature work or bug fixes
- Coordinating changes across multiple repositories or monorepo subprojects
- Generating draft PRs from a high-level goal and running integrated verification tests
- Running repeatable builds that produce artifacts and per-issue logs
What's included
- Scripts: repository contains control-plane and node registration examples (start af, python -m swe_af) and example curl triggers for build/plan/execute endpoints
- References: in-repo docs (ARCHITECTURE.md, deployment.md) explain architecture and artifact layout
- Instructions: README-style usage with endpoint examples, configuration keys (runtime, models), role definitions, and artifact locations
Compatible agents
Inferred support for model-backed coding runtimes and CLI-driven backends such as Claude Code, OpenCode-style providers, and generic LLM coding agents (e.g., Codex/GLM/Gemini-like toolchains).
Not yet audited
This skill has not been reviewed by our automated audit pipeline yet.