Agent Swarm
Open-source, self-hosted AI company operating system where a lead agent delegates work to specialized worker agents with shared memory.
At a Glance
About Agent Swarm
Agent Swarm is an open-source, self-hosted operating system for AI work built by Desplega Labs and released under the MIT license. A lead agent receives requests from channels such as Slack, GitHub, GitLab, Linear, email, or the API, breaks them into tasks, and routes them to worker agents running in isolated Docker containers. The latest release listed is v2.6.0.
What It Is
Agent Swarm coordinates multiple AI agents as a team. The lead agent plans and delegates, while workers run on a harness of your choice, such as Claude Code, Codex, pi, OpenCode, Cursor, Amp, Grok, Devin, or ACP agents. Memory and identity persist across sessions, so what one agent learns is available to the others. Workflows, schedules, scripts, and apps support recurring work, and review gates let people approve outputs such as pull requests.
How Work Flows
A request arrives, for example a bug report in Slack. The lead routes it to a worker, which might open a pull request. A reviewer agent and automated test steps can run in parallel, deterministic steps run the same way each time, and a person approves anything that needs approval. The product also documents deferred tasks that wake when watched tasks finish or a deadline arrives, and schema-validated task results for callers needing structured JSON.
Agents, Templates, and Models
The site describes 11 official agent templates, including Lead, Coder, Reviewer, Researcher, Tester, Content Writer, and Forward Deployed Engineer. Each agent can be assigned its own harness and model, so different model providers can be mixed in one swarm. Custom agents can be created from a template or from scratch.
Setup Path
Installation is done by asking a coding agent to read the operator skill, by running a one-line installer that requires Docker, or by using the provided Docker Compose example. A Helm chart is available for Kubernetes. Each deployment exposes its own REST API and MCP server, with a dashboard for setup and monitoring. A CLI is published on npm as @desplega.ai/agent-swarm.
Community Discussions
Be the first to start a conversation about Agent Swarm
Share your experience with Agent Swarm, ask questions, or help others learn from your insights.
Pricing
Open source
MIT licensed, self-hosted on your own infrastructure; you pay for your own infrastructure and model usage.
- Full source on GitHub (MIT)
- Run anywhere Docker runs
- BYO model keys, BYO models
- Air-gapped if you need it
- Community support on Discord
Enterprise
Support for your deployment; services, SLA terms and price are agreed with your team via a scoped quote.
- Deployment and onboarding support
- Priority support and agreed SLAs
- SSO and SAML integration
- Custom models and provider integration
- Custom integrations and MCP servers
Capabilities
Key Features
- Lead agent that delegates tasks to worker agents
- Workers in isolated Docker containers
- Persistent shared memory and identity with vector search
- Choice of harness and model per agent
- Workflows, schedules, scripts, and apps for recurring work
- Human-in-the-loop approval and review gates
- Schema-validated task results
- Deferred tasks
- Extensions for trusted TypeScript hooks
- Realtime rooms
- REST API and MCP server per deployment
- Dashboard with guided setup
- Agent templates
- Kubernetes Helm chart and Docker Compose deployment