Quick Start¶
Dev Agents needs three things: an LLM API key, a local git checkout to analyze, and at least one entrypoint to talk to it through. Everything else is optional.
1. Configuration¶
Generate a .env with the Setup Wizard, or start from the example:
The minimum:
ANTHROPIC_API_KEY=sk-ant-... # or OPENAI_API_KEY / AWS credentials
LLM_MODEL_LARGE=anthropic:claude-sonnet-5
LLM_MODEL_SMALL=anthropic:claude-haiku-4-5
GIT_REPO_PATH=/code # path to the checkout to analyze
Model strings follow Pydantic AI (provider:model-name).
Add Slack tokens and provider credentials as needed — see
environment variables.
2. Run¶
Add -v (after the image name) for verbose logs.
3. Talk to it¶
On startup, Dev Agents detects which services are configured and starts them all in parallel: Slack, platform services, NATS, the HTTP entrypoints (AG-UI and the OpenAI-compatible API), and — when stdin is a terminal — an interactive CLI chat.
dev-agents # interactive chat + every configured service
dev-agents --prompt "How does the auth flow work?" # single prompt, then exit
dev-agents -v # verbose logging
In Slack:
Next steps¶
- Environment variables — the full reference
- config.yaml — projects, providers, agents
- Slack — app manifest and scopes
- Hooks — extend the agent with your own tools