SWE-AF
Autonomous software engineering fleet of AI agents for production-grade PRs on AgentField: plan, code, test, and ship.
SWE-AF (Autonomous Engineering Team Runtime) is a multi-agent AI system built on AgentField that orchestrates complete software engineering workflows from a single API call. It deploys specialized agent roles—product managers, architects, coders, reviewers, and testers—that collaboratively plan, implement, test, and ship production-grade code changes. The system supports multiple LLM providers (Claude, OpenRouter, OpenAI, Google) with per-role model assignment, runs agents in parallel across isolated git worktrees, and includes adaptive control loops that replan work when issues prove harder than expected. Developers trigger builds via CLI, HTTP API, or Railway deployment, receiving merged branches with passing tests and optional pull requests.
Key Features
Use Cases
- 01Autonomous feature development: describe a goal and receive a complete, tested implementation with pull request
- 02Refactoring and hardening existing code across authentication, billing, or API layers
- 03Multi-repository feature coordination spanning primary applications and shared library dependencies
- 04Delegating well-scoped coding tasks from parent harnesses (Claude Code, Codex) to cheaper models via implement_issue
- 05Production-grade code generation with comprehensive test coverage (99%+ measured in benchmarks)
- 06Automated bug fixes with CI validation and autonomous repair of failing checks
Related Agents
View morehermes-agent
The agent that grows with you
agency-agents
A complete AI agency at your fingertips - From frontend wizards to Reddit community ninjas, from whimsy injectors to reality checkers. Each agent is a specialized expert with personality, processes, and proven deliverables.
openinterpreter
A coding agent for open models like Kimi K3
cline
Autonomous coding agent as an SDK, IDE extension, or CLI assistant.
SWE-AF — FAQ
What is SWE-AF?+
SWE-AF is an autonomous multi-agent system built on AgentField that orchestrates full software engineering workflows—planning, coding, testing, and shipping—from a single API call or CLI command.
How do I install SWE-AF?+
Install via AgentField control plane with 'af install https://github.com/Agent-Field/SWE-AF', deploy with one click on Railway, or run locally with Docker Compose. All methods require one LLM provider API key (ANTHROPIC_API_KEY or OPENROUTER_API_KEY recommended).
Which AI clients and models does SWE-AF support?+
SWE-AF works with Claude (via Anthropic API or Claude Code CLI), OpenRouter (200+ models), OpenAI, Google, and MiniMax. You can assign different models to different agent roles (e.g., opus for coding, haiku for QA) in a single build.
What API keys or prerequisites are required?+
Exactly one LLM provider key is required: OPENROUTER_API_KEY (simplest, accesses 200+ models), ANTHROPIC_API_KEY, CLAUDE_CODE_OAUTH_TOKEN, OPENAI_API_KEY, or GOOGLE_API_KEY. A GH_TOKEN (GitHub personal access token with repo scope) is needed only for cloning private repositories and opening pull requests.
Is SWE-AF free to use?+
SWE-AF itself is open-source (Apache 2.0 license), but you pay for the underlying LLM API calls. Benchmark example: a 10-issue Node.js CLI app with full tests cost $19.23 with Claude haiku or $6 with MiniMax M2.5.
Can I use SWE-AF as a sub-harness from another coding agent?+
Yes, the implement_issue entry point accepts fully-scoped issues from parent harnesses (Claude Code, Codex, OpenCode) and returns just the coding loop on an isolated branch, skipping all planning agents. Typical cost is 4-8 LLM calls per issue.
How do I install SWE-AF?+
Open the source repository on GitHub and follow its README. SWE-AF is a agent — MCP Agents Market links you directly to the official repo.
Is SWE-AF free?+
SWE-AF is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.