OpenSwarm

OpenSwarm is an open-source multi-agent AI team you can run locally (terminal UI) or deploy as an API.

Instead of a single "do-everything" bot, OpenSwarm routes each request to the right specialist agent (research, data analysis, docs, slides, images, video, and a virtual assistant) and coordinates the result for you.

Built on Agency Swarm and the OpenAI Agents SDK.

What can I do with it?

Examples you can paste into the terminal UI:

• "Research the current competitive landscape for X, cite sources, and summarize in bullets."
• "Analyze mnt/data.csv, find anomalies, and produce a chart plus a short narrative."
• "Create a 6-slide pitch deck about X with a modern theme, then export to PPTX."
• "Draft a professional one-page memo about X and export to PDF."
• "Generate a hero image for my landing page, then create 3 variants."
• "Combine these clips, trim pauses, add captions, and export a final video."

OpenSwarm can also connect to external services (Gmail, Slack, GitHub, etc.) via Composio if you set the optional keys.

Agent Roster

| Agent | What it does | |---|---| | Orchestrator | Routes every user request to the right specialist(s). Never answers directly — pure coordination. | | Virtual Assistant | Handles everyday tasks: writing, scheduling, messaging, task management. Gains 10,000+ external integrations via Composio (Gmail, Slack, GitHub, HubSpot, and more). | | Deep Research | Conducts comprehensive, evidence-based web research with citations and balanced analysis. | | Data Analyst | Analyses structured data, builds charts, runs statistical models — all inside an isolated IPython kernel. | | Slides Agent | Generates complete, visually polished HTML slide decks, then exports them to PPTX. | | Docs Agent | Creates formatted Word documents and PDFs from outlines or raw content. | | Image Generation Agent | Generates and edits images using Gemini 2.5 Flash Image / Gemini 3 Pro Image and fal.ai. | | Video Generation Agent | Produces videos via Sora (OpenAI), Veo (Google), and Seedance (fal.ai); also edits and combines clips. |

Prerequisites

| Requirement | Notes | |---|---| | Python | 3.10 or newer | | Node.js | 20 or newer |

Everything else (Python packages, Node.js dependencies, Playwright / Chromium, LibreOffice, Poppler) is installed automatically on first run.

Quick Start

Option A — Local install (interactive terminal)

git clone https://github.com/VRSEN/openswarm.git
cd openswarm
python run.py

On first run, a setup wizard walks you through choosing a provider and entering your API keys. Everything else (Python packages, Node.js dependencies, Playwright browser) installs automatically.

Once configured, python run.py drops you straight into the terminal UI. Outputs and generated files are written to ./mnt/.

Option B — Docker (API server; no system dependencies required)

Requires Docker.

git clone https://github.com/VRSEN/openswarm.git
cd openswarm
cp .env.example .env        # fill in at least one provider key
docker-compose up --build

The FastAPI server starts on http://localhost:8080.

Running as an API

python server.py        # local install
# or
docker-compose up       # Docker

Starts a FastAPI server on port 8080. The agency is exposed at /open-swarm.

Configuration

Provider & model

Set your provider key and optionally override the default model in .env:

OPENAI_API_KEY=sk-...

# Optional: override the model used by all agents
DEFAULT_MODEL=gpt-4o

The onboarding wizard sets these automatically. You can also edit .env directly at any time — see .env.example for all options.

Optional integrations

| Key(s) | Feature unlocked | |---|---| | ANTHROPIC_API_KEY | Claude models via LiteLLM | | COMPOSIO_API_KEY + COMPOSIO_USER_ID | 10,000+ external integrations via Virtual Assistant | | SEARCH_API_KEY | Web search, Scholar search, Product search (searchapi.io) | | GOOGLE_API_KEY | Gemini image generation/editing, Veo video generation | | FAL_KEY | Seedance video generation, video editing, background removal | | PEXELS_API_KEY | Pexels stock photo search (Slides Agent) | | PIXABAY_API_KEY | Pixabay stock photo search (Slides Agent) | | UNSPLASH_ACCESS_KEY | Unsplash stock photo search (Slides Agent) |

Tools that require a missing key raise a clear error at call time — the agent will tell you what to add.

Adding a New Agent

1. Create a new folder: my_agent/
2. Add my_agent.py (Agent definition), instructions.md, and tools/
3. Register the agent in swarm.py → create_agency()

See the Agency Swarm docs for the full agent creation guide.

Learn More

| Component | Library | |---|---| | Multi-agent framework | Agency Swarm | | LLM runtime | OpenAI Agents SDK | | External integrations | Composio | | Multi-provider LLM routing | LiteLLM | | API deployment | FastAPI + Uvicorn | | SaaS platform | Agent Swarm |

License

MIT — see LICENSE.