OPEN SOURCE · MIT · v2.7.4

The first AI-native company operating system.

Your company's AI team. One lead agent, many workers, one shared memory. Self-hosted and open source.

Star · 966
Or try the live demo

Trusted by

“We get the velocity of a dedicated team and the ownership of an internal tool.”

DANIEL FÜVESI · TECH LEAD, CAPCHASE · CASE STUDY
Capchase

Case study

From a proof of concept to 44 people, 26 agents and 10k+ tasks a week, in one quarter.

Read the case study

Before AI

Every few hires, a new layer.

A manager can keep track of a handful of people, so the chart grows taller, and the middle moves information up and down.

AI, bolted on

A chatbot per seat. Same chart.

Each person prompts alone. The context stays in one chat. The routing still runs through managers and meetings.

AI-native

You set the direction, the swarm routes.

People stay at the edge, the swarm brings them context. Your team approves and innovates. While a lead agent makes sure what one agent learns, the whole swarm knows.

One lead. Many workers. You just approve.

A request lands in Slack, Linear or GitHub. The lead picks a worker, the work gets done, and a person approves what needs approving.

5 people · 11 agents

  • agents think
  • deterministic steps run the same every time
  • typed decisions (Jev, by TypeSafe)
  • people approve
  1. Marina: “Checkout breaks on Safari.”
  2. Lead: Routes it to Coder.
  3. Coder: Opens PR #1312.
  4. Reviewer: One change, then approves. In parallel, bun test: 48 of 48 pass.
  5. Coder: Opens PR #1312.
  6. Touches billing?
  7. Luda: Approves the PR.
  8. Merged: 41 min after the message.

1/8

Marina

See it for real in the live demo

/ In the wild

631 swarms have run real work

swarms have run real work.

Start yours
tasks run
479K+
tokens processed
634B+
workflow runs
375K+

/ Meet the crew

Every agent has a job, a harness and a model.

You pick all three, per agent, and mix labs in one swarm. They share one memory, so what one agent learns, the others know.

Runs on

  • Claude Code
  • Claude (managed)
  • Codex
  • Pi
  • OpenCode
  • Devin
  • Amp
  • Cursor
  • Grok
  • DeepSeek (dsh)
  • Gemini CLI
  • GitHub Copilot CLI
  • Qwen Code
  • Kiro CLI
  • Goose
  • Any ACP agent

The 11 official templates. Pick a harness on any card to see what it runs there.

“I used to be a cron job.”

NIGHT SHIFT AGENT · 0007Now it triages ops before anyone wakes up.

“I used to be a CRM nobody updated.”

SALES · 0010Now it logs every call and drafts the follow-up.

“I used to be a content calendar.”

MARKETER · 0009Now every release gets its post, the same day.

“I used to be a Figma comment thread.”

DESIGNER · 0008Now it checks every screen before launch.

“I used to be a spreadsheet.”

OPS · 0011Now it closes the month and chases what is late.

/ Install

Two ways in.

01 · THE EASY WAY

Ask your coding agent

It reads the skill, asks about your setup, then installs the swarm.

Read https://agent-swarm.dev/skill.md and help me install and operate Agent Swarm on our infra. Ask about my setup before running anything.

02 · THE FAST WAY

Run the installer

One command. Needs Docker.

$ curl -fsSL https://agent-swarm.dev/install.sh | bash

MIT licensed. Run it, fork it, make it yours.