Agent1 is an autonomous GitHub engineering partner with deterministic workflows and safety-first controls.
Value Proposition • Quick Start • Contributing • License
Agent1 collapses issue and PR lifecycle handling, policy-governed execution, and operator-grade transparency into one cohesive system with durable state, deterministic transitions, and strict environment isolation.
Run backend and dashboard together from repository root with one Docker command. Once started, your frontend is exposed on:
http://localhost:8080/ — main dashboard UI http://localhost:8080/ — same app (nginx falls back to index.html)
Frontend API proxy endpoint:
http://localhost:8080/api/* — proxied to backend http://backend:$PORT/* inside compose (PORT default 8000)
Useful examples through the frontend: http://localhost:8080/api/health -> backend /health http://localhost:8080/api/docs -> backend /docs http://localhost:8080/api/openapi.json -> backend /openapi.json
Environment resolution for Docker is explicit:
- Compose reads values from shell environment first.
- Missing values fall back to
.env. - A complete starter template is in
.env.example.
activemode (currently supported)
AGENT1_DOCKER_MODE=active docker compose up --build- Backend API:
http://localhost:${PORT:-8000} - Backend health:
http://localhost:${PORT:-8000}/health - Backend API docs:
http://localhost:${PORT:-8000}/docs - Dashboard:
http://localhost:8080 - Dashboard-to-backend API proxy:
http://localhost:8080/api/* - Runtime behavior:
prodenvironment +activemode. - Uses runtime credentials and database settings from
.env. GITHUB_TOKENmust be set for Agent1 runtime.- Fork note: set
GITHUB_USERto your bot account and aligncontrols/policies/default.json(agent_actorandmutating_credential_owner_by_environment) to that same account. - Full variable-by-variable guidance:
docs/Developer/Environment-Variables.md. - Startup remains a single command; migration failures now fail closed (no automatic Alembic stamping).
- If startup fails on migrations, run:
docker compose run --rm --entrypoint sh backend -lc 'alembic upgrade head'- then rerun
AGENT1_DOCKER_MODE=active docker compose up --build.
devmode (currently unsupported)
AGENT1_DOCKER_MODE=devis intentionally blocked by runtime entrypoint safeguards.- This mode will be re-enabled after adding a dedicated dev runtime user/token and additional isolation controls to keep dev and active behavior strictly separated.
Stop the stack:
docker compose downFor partner-facing usage flow, see User Docs.
For architecture and implementation baseline, see Architecture and Developer Docs.
The simplest way to contribute is by joining open discussions or picking up an issue:
Before contributing, start with Developer Docs.
Report vulnerabilities privately through GitHub Security Advisories.
If you use Agent1 for published work, please cite:
Agent1 [Computer software]. (2026). Retrieved from http://github.com/vaquum/agent1.
