Shannon
A production-oriented multi-agent orchestration framework.
What is Shannon?
A production-oriented multi-agent orchestration framework.
Overview of Shannon
What it is
Shannon is a production-oriented multi-agent orchestration framework. It is designed to ship reliable AI agents to production, providing multi-strategy orchestration, swarm collaboration, token budget control, human approval workflows, and time-travel debugging. The project is licensed under MIT and is open source.
Key features
- Multi-strategy orchestration: tasks are automatically routed based on complexity, supporting Simple, DAG, ReAct, Research, Exploratory, Browser Use, Domain Analysis, and Swarm strategies.
- Token budget control: hard token budgets per task/agent with automatic model fallback.
- Human-in-the-loop approval: enable approval gates via OPA policy or workflow templates with require_approval: true.
- Time-travel debugging: replay any execution step-by-step using Temporal workflows.
- Real-time observability: event streaming, Prometheus metrics, OpenTelemetry tracing.
- WASI sandbox for code execution, OPA policies, multi-tenant isolation.
- Supports 10+ LLM providers: OpenAI, Anthropic, Google, DeepSeek, xAI, MiniMax, Groq, Qwen, Meta, Zhipu, Kimi, and local models via Ollama, LM Studio, vLLM.
Use cases
Shannon is suitable for production AI agent deployments requiring reliable orchestration, cost control, and observability. Use cases include multi-step research workflows, swarm multi-agent collaboration, scheduled tasks, session continuity for conversational agents, and tasks requiring human approval gates. It can be used for code review, market analysis, web browsing, and domain-specific deep research.
Getting started
Prerequisites: Docker and Docker Compose, and an API key for at least one LLM provider.One-command install:bashcurl -fsSL https://raw.githubusercontent.com/Kocoro-lab/Shannon/main/scripts/install.sh | bashSee the repository README for platform-specific guides and building from source.
License and pricing
The project is licensed under MIT and is open source. Pricing is not stated beyond being open source.
Based on the project's public README and repository metadata.