An autonomous AI engineering agent that works on GitHub issues in sandboxed VMs.

aieng is an autonomous AI engineering agent that uses Claude API inside isolated virtual machines to autonomously work on GitHub issues.

Workflow:

1. Issue Intake - Accept a GitHub issue as input
2. Implement & PR - Clone repo, implement solution with tests, open PR
3. Review Loop - Monitor PR comments, make fixes or ask for clarification

Components:

• VM Management - Abstraction layer for virtualization (Tart and Docker providers)
• Orchestrator - Manages VM lifecycle, IssueWatcher polling, and VM pool
• Agent - Runs inside VMs, uses Claude API for autonomous implementation

# Watch for GitHub issues with reactions - VMs auto-start/stop as needed
aieng --repo owner/repo

# Specify image for VM creation on first run
aieng --repo owner/repo --image base-image

# Use Docker provider instead of Tart
aieng --provider docker --repo owner/repo

# Process multiple issues in parallel with VM pool
aieng --repo owner/repo --max-vms 3

# Debug commands for VM troubleshooting
aieng dbg list                          # List all VMs
aieng dbg shell owner-repo-worker       # Open SSH session
aieng dbg exec owner-repo-worker "ls"   # Execute command

See doc/operations.md for installation, setup, and detailed usage.

MIT