AgentRail Docs

Dashboard & Hosted Runner

The primary flow — sign in, connect GitHub, queue issues, and run a local worker.

The hosted dashboard + self-hosted runner is the primary and required flow. Sign up at app.useagentrail.com, connect your GitHub repos, then run agentrail login and agentrail runner on your machine. The runner claims queued issues over HTTP, executes them locally with your own LLM keys, opens PRs, and reports back to the dashboard.

Sign in

agentrail login

login signs this machine into an AgentRail workspace via an OAuth device flow, saving credentials to ~/.agentrail/credentials.json. Use --url BASE_URL to point at a self-hosted server. Check who you're signed in as with whoami, and sign out with logout:

agentrail whoami
agentrail logout

Login is required for all CLI commands. The only exempt commands are login, logout, whoami, and --help.

Connect GitHub

On the dashboard Connectors page, authorise GitHub and select the repos you want AgentRail to watch. Configure:

  • Trigger labelready-for-agent (issues with this label are enqueued)
  • Poll interval — how often the dashboard checks for newly labelled issues

Queue issues in the console

Once a repo is connected, queue work by labelling a GitHub issue ready-for-agent. The issue enters the workspace Issue Queue, carrying its tier (which model), remaining budget, and state — and leaving in exactly one Run Outcome terminal: Green, Escalated-to-human, or Blocked.

You can also open the interactive console directly:

agentrail console

Run the local worker

agentrail runner --concurrency 2

runner is the local worker that polls the workspace queue, claims available issues, and executes them on your machine. The work happens locally — your keys, your sandbox — while the workspace coordinates the queue and stores run outcomes.

FlagDescription
--idle SECONDSHow long to wait for work before idling out.
--onceProcess one batch, then exit.
--concurrency NNumber of issues to run in parallel.

What the dashboard adds

  • Shared run history — context packs, cost, gate evidence, and outcomes from all your runs in one place across the team.
  • Cost across repos and teams — real-dollar cost rolled up beyond a single laptop, so spend never surprises you.
  • A source-linked audit trail — who or what performed each sensitive action.
  • Members and roles — invite teammates and govern who can configure review gates and manage repositories.

Indexing still stays local on each developer's machine; the dashboard stores run metadata, context-pack citations, costs, and audit events — not your source by default.