Rapid agent swarms. Make any agentic workflow exponentially faster.
Funded by Y Combinator
Get started
Run your first agent swarm.
A swarm splits your task across coding agents, each in its own sandbox. They work in parallel, then merge into one patch or pull request.
npm install -g @acyclic-labs/cli macOS and Linux. Installs the acyclic command. No Node? curl -fsSL https://acyclic.dev/install.sh | sh fetches the binary with a verified checksum
(read it first).
1. Sign in
Opens your browser to approve this device. One login lasts 90 days.
acyclic swarm login --site https://acyclic.dev
2. Describe the task
Run it inside your project:
--dir .uploads your working files (no history, no secrets).--watchfollows the agents live.acyclic swarm run --task "fix the flaky test in tests/api" --dir . --watch
3. Take the result
Apply the merged patch to your checkout, or open a pull request from the dashboard.
acyclic swarm pull <swarm-id> --apply
Prefer not to install anything? Start a swarm from the dashboard or connect a coding agent over MCP. Full reference: CLI guide.
Talk to us
Scale agent swarms to tasks that are too big for one agent.
Acyclic runs large tasks as a swarm: fan work out across many agents, fork the ones worth exploring, watch every branch as it runs, and keep only the results you verified.
We're onboarding teams gradually, starting with those that already run agents in volume. Tell us what you're running and one of the founders will set up a short conversation about scaling it.
- Startups building domain-specific agents that need to run at scale
- Companies already operating swarms of agents in production
- Teams running many Claude Code or Codex sessions on large codebases