Files
aiworker-agent/README.md
Hector Ros 1db98be556
Some checks failed
Build and Push Agent / build (push) Failing after 50s
Initial agent implementation
- Node.js Alpine base with Claude Code CLI
- MCP configuration for backend communication
- Git and development tools pre-installed
- K8s deployment manifests
- CI/CD workflow for automatic builds

Co-Authored-By: Claude Sonnet 4.5 (1M context) <noreply@anthropic.com>
2026-01-20 01:23:30 +01:00

48 lines
1.3 KiB
Markdown

# AiWorker Agent
Claude Code agent running in Kubernetes pods to autonomously complete development tasks.
## Architecture
The agent:
- Runs in K8s namespace `agents`
- Communicates with backend via MCP protocol
- Has access to Git repositories via Gitea
- Reports progress and status to backend API
## Local Development
This is a containerized agent meant to run in Kubernetes. For local testing:
```bash
docker build -t aiworker-agent:local .
docker run -it --rm \
-e ANTHROPIC_API_KEY=your_key \
-e BACKEND_URL=http://localhost:3000 \
aiworker-agent:local \
/bin/bash
```
## Deployment
See `k8s/agents/README.md` for deployment instructions.
## MCP Tools
The agent uses these MCP tools to communicate with the backend:
- `get_next_task` - Get next task from queue
- `update_task_status` - Update task state
- `create_branch` - Create Git branch for task
- `create_pull_request` - Create PR when task is complete
- `ask_user_question` - Request user input when needed
## Environment Variables
- `ANTHROPIC_API_KEY` - Claude API key
- `BACKEND_URL` - Backend API URL
- `MCP_SERVER_URL` - MCP server URL
- `GITEA_URL` - Gitea URL
- `GITEA_TOKEN` - Gitea access token
- `POD_NAME` - Pod name (auto-injected by K8s)
- `NAMESPACE` - Namespace (auto-injected by K8s)