BullswarmRoute coding work by quota. Verify it by content.

Put your logged-in coding-agent subscriptions to work as one paced, observable fleet — without turning them into an API proxy.

Get started

Read the playbook

CLI reference

⭐ Star on GitHub

Pacing router

Spend subscription quota before it expires. Weekly and monthly pace, reset urgency, five-hour headroom, and current load all shape the pick.

See how routing works

Lanes

Describe the work as analyze, build, or chore. Eligible pools compete at runtime instead of being hard-wired to a task type.

Learn the vocabulary

Workflows

Author a dependency graph, let ready steps fan out across providers, then bring the results back together in one shared workspace.

Build a workflow

Verification

Exit code is not the verdict. Bullswarm judges saved output by content, and workflow requirements can require evidence another worker inspected.

Read the result envelope

Dashboard

Follow Home, Runs, Run, Step, Budget, Stats, and Fleet from one keyboard-and-mouse dashboard while work continues independently.

Tour the dashboard

Meters

Provider-owned usage readers supply quota windows and resets. Stale or unavailable measurements stay visibly unknown instead of becoming zero.

Understand usage and cost

Packaged skill

Install Bullswarm's guidance into Claude Code, Codex, or Grok so your main agent can choose a bounded run or author a workflow for you.

Connect an agent CLI

Claude Code Mod

Open a read-only Bullswarm pane inside Claude Code with the same run, step, and usage facts as the terminal dashboard.

Open the Claude integration

One fleet, your subscriptions

Bullswarm drives the official headless CLIs already installed and logged in on your machine. Start with one bounded task, or give your main agent a goal and a program that can spread independent work across the fleet.

Install and run your first task or read the day-to-day playbook.