Autonomous Issue Agent

Autonomous Issue Agent (AIA)

An autonomous agent that implements GitHub Issues using Claude Code.

**Works with any GitHub repository** — not limited to a specific tech stack or project type.

Implementation: Python Agent

**Recommended Implementation:** Standalone Python agent (headless subprocess)

The Python-based autonomous agent runs as a background service:

• ✅ Fully automated polling loop - Monitors multiple repositories 24/7
• ✅ Headless operation - Runs without interactive Claude Code session
• ✅ Multi-repository support - Process issues across different repos
• ✅ Session management - Resumes work across restarts
• ✅ Built-in tools - Semantic search and smart test runner
• ⚠️ No MCP support - MCP causes hangs in subprocess headless mode
• ℹ️ Token usage - ~15-25k tokens per issue

**Note:** An experimental plugin implementation exists in `plugin/` but has stability issues with MCP in headless mode and is not actively maintained. The Python agent is the stable, production-ready implementation.

---

🛠️ AI-Powered Development Tools

Looking for standalone AI development tools? Check out [**python-dev-tools**](https://github.com/aignermax/python-dev-tools):

• Semantic Search - Natural language code search using AI embeddings
• Smart Test Runner - Automatically finds and runs only relevant tests

These tools can be used independently or integrated with this agent for enhanced capabilities.

---

How it works

GitHub Issues (label: agent-task) across multiple repos
        │
        ▼
   Agent polls every 15s (configurable)
        │
        ▼
   For each repository:
     - Check for agent-task issues
     - Clone/pull latest changes
     - Create feature branch
        │
        ▼
   Claude Code (headless):
   - Reads the entire repository
   - Understands the architecture
   -