Repolode
Submit
Kocoro-lab

Kocoro-lab/Shannon

A Go, Rust and Python multi-agent orchestration framework for running LLM agents in production, with Temporal workflows and token budgets.

View on GitHubWebsite
Stars
2,316
Forks
361
Open issues
8
Last commit
2 months ago

What is it?

Shannon is a multi-agent orchestration framework aimed at running AI agents reliably in production. Tasks are submitted through a REST API, an OpenAI-compatible endpoint, a Python SDK or a desktop app, and are routed to one of several execution strategies such as simple single-agent responses, DAG fan-out/fan-in, ReAct loops, research, tree-of-thoughts exploration and swarm teams, based on task complexity.

The system runs workflows on Temporal for fault tolerance and offers replay of past executions for debugging. Token budgets can be set per task, with automatic fallback to cheaper models as limits approach. Code execution happens in a WASI sandbox, and optional human approval gates can pause a workflow until a decision is submitted. Metrics and tracing use Prometheus and OpenTelemetry.

Shannon supports Anthropic, OpenAI, Google, DeepSeek, xAI, Groq and local models through Ollama, LM Studio or vLLM. It is deployed with Docker Compose and includes scheduled tasks, session continuity and skills that can be defined in YAML and Markdown.

Key features

  • 01Multiple execution strategies chosen automatically by task complexity
  • 02Temporal workflows with replay of past executions for debugging
  • 03Per-task token budgets with automatic model fallback
  • 04WASI sandbox for code execution and optional human approval gates
  • 05OpenAI-compatible API endpoint, Python SDK and desktop app
  • 06Support for many LLM providers, including local models via Ollama and vLLM

Use cases

  • Running multi-step research or analysis tasks with cost limits
  • Adding approval checkpoints before agents take sensitive actions
  • Pointing existing OpenAI SDK code at a self-hosted agent backend
  • Scheduling recurring agent tasks such as daily market analysis

Quick start

  1. 1

    Run the one-command install script from the README (review it first; it pipes to bash)

  2. 2

    Install the Python SDK

    pip install shannon-sdk
  3. 3

    Submit your first task via the REST API

    curl -X POST http://localhost:8080/api/v1/tasks -H "Content-Type: application/json" -d '{"query": "What is the capital of France?", "session_id": "demo"}'

Things to know

  • Requires Docker and Docker Compose plus an API key from at least one LLM provider.
  • The one-command install runs a remote script piped into bash.
  • The optional Playwright browser service adds a roughly 3.4GB image.

In collections

Multi-agent orchestration frameworksOpen-source AI agent frameworks & apps

Topics

#agent#multi-agent#llm#orchestration#temporal#workflow#wasi#golang#ai#multi-agent-systems

FAQ

What is Shannon?
A Go, Rust and Python multi-agent orchestration framework for running LLM agents in production, with Temporal workflows and token budgets. Shannon is a multi-agent orchestration framework aimed at running AI agents reliably in production. Tasks are submitted through a REST API, an OpenAI-compatible endpoint, a Python SDK or a desktop app, and are routed to one of several execution strategies such as simple single-agent responses, DAG fan-out/fan-in, ReAct loops, research, tree-of-thoughts exploration and swarm teams, based on task complexity.
Is Shannon free to use? What license does it have?
Shannon is open source under the MIT license. Check the license terms before using it in commercial products.
Is Shannon still maintained?
Updated within 6 months. The last commit to Shannon was on Sep 5, 2026.
What can I use Shannon for?
Running multi-step research or analysis tasks with cost limits Adding approval checkpoints before agents take sensitive actions Pointing existing OpenAI SDK code at a self-hosted agent backend Scheduling recurring agent tasks such as daily market analysis
How do I install Shannon?
Run the one-command install script from the README (review it first; it pipes to bash) Install the Python SDK: pip install shannon-sdk Submit your first task via the REST API: curl -X POST http://localhost:8080/api/v1/tasks -H "Content-Type: application/json" -d '{"query": "What is the capital of France?", "session_id": "demo"}'
Where can I get Shannon?
The source code and installation instructions are on GitHub: https://github.com/Kocoro-lab/Shannon

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects