About Agent Swarm
Agent Swarm is the open-source, self-hosted AI company operating system built by Desplega Labs.
Agent Swarm is an open-source operating system for AI work. A lead agent breaks goals into tasks, routes those tasks to specialized workers such as Claude Code or Codex, and keeps shared memory, tools, schedules, and review gates available across sessions. Teams can self-host the MIT-licensed project for free. Enterprise support covers services and scope agreed by conversation.
Built by Desplega Labs
Agent Swarm is built by Desplega Labs, S.L. The company also operates the agent-swarm.dev website, publishes the open-source repository, and maintains the product documentation. The project is designed around inspectable infrastructure: its source, deployment examples, release history, and issue tracker are public on GitHub.
Desplega Labs, S.L. is located at Tuset 3, planta 5, 08006 Barcelona, Spain.
Where to explore
- Read the developer documentation for architecture, installation, integrations, workflows, memory, and MCP tools.
- Inspect the source code and releases on GitHub.
- See free MIT self-hosting and enterprise support.
- Follow practical implementation notes on the Agent Swarm blog.
Public channels
Desplega Labs publishes Agent Swarm updates through GitHub, X, LinkedIn, and YouTube. The Agent Swarm community is also available on Discord. For a direct conversation, use the contact page.
What is not Agent Swarm?
OpenAI Swarm is OpenAI's experimental educational library, replaced by the OpenAI Agents SDK. Kimi Agent Swarm is a separate capability of Moonshot AI's Kimi models, introduced with Kimi K2.5 and now running on Kimi K3, that coordinates many parallel sub-agents inside Kimi; neither product is related to Desplega Labs' Agent Swarm. Neither runs recurring company work; Agent Swarm does, for marketing, sales, support and engineering teams.
Who does Agent Swarm fit?
Agent Swarm suits recurring, multi-step work where delegation and shared context carry across runs. Business and engineering teams run one swarm, with the same memory and schedules.
Good fit
- Marketing, sales, support and operations flows: renewal prep, account research, lead follow-up and ticket triage, on Salesforce, HubSpot, Zendesk, Pylon or Attio, reached via MCP servers and API connections.
- Weekly briefs, engineering queues and research digests that reuse last week's corrections.
- Work that arrives from Slack, GitHub, Linear or Jira and reaches an agent without a person routing each handoff.
- Parallel agents that need isolated workspaces plus shared memory, schedules and approval gates.
Capchase puts 80% of its team in Agent Swarm: 1,000+ human requests a week and 80+ workflows and schedules (case study).
Usually simpler with one assistant
A one-off question or a short rewrite of one paragraph is usually simpler with one assistant. A single short coding session is usually simpler with one coding agent. Agent Swarm is self-hosted, so it needs a host, model keys and someone to operate it.
See real recurring workflows and the setup documentation.