Skip to content

earlvanze/sage-router

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1,542 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Sage Router

A local-first, OpenAI-compatible router for AI agents.

Use one endpoint for the model access you authorize: local Ollama, provider accounts, and hosted Sage Router capacity. The router selects eligible routes, keeps failures scoped to the affected credential or model, and exposes what happened.

Hosted API | Quickstart | Getting started | Integrations

Start here

Supported access

Sage Router is BYOK and subscription-aware. It routes access you authorize; hosted plans provide routing and operational features, not pooled personal provider accounts.

Access you have Connect it with Route prefix
ChatGPT / Codex subscription Active Codex OAuth profile openai-codex/*
Ollama Cloud subscription Authenticated local Ollama runtime ollama-cloud/*
Z.ai GLM Coding Plan Personal Coding Plan API key zai-coding/*
Anthropic, OpenAI, Gemini, Vertex, OpenRouter, NVIDIA NIM, xAI, Mistral, Groq, and compatible APIs Your API key or workload credentials Provider-specific prefix
Local Ollama and private OpenAI-compatible endpoints Local endpoint and, if required, its credential Your configured provider

Start with the provider integration guides. For Z.ai, use the dedicated Coding Plan setup, not its general API endpoint.

What it does

  • Routes OpenAI- and Anthropic-compatible requests across eligible cloud and local models.
  • Keeps authentication, quota, and model failures isolated to the affected route so healthy alternatives remain available.
  • Supports streaming, Responses background tasks, tool calls, images, and common coding-agent harnesses.
  • Runs Fusion for bounded planning, parallel worker execution, and final verification.

Documentation

Need Start here
First hosted or self-hosted request Getting started
Codex, OpenClaw, and other tools Integration guides
Docker, Umbrel, Tailnet, or GCP Deployment map
Route profiles router-profiles.json
Fusion design Fusion
Hosted edge and billing operations Deployment map

Contributing

Read SECURITY.md and CONTRIBUTING.md before changing runtime behavior. Do not commit credentials, OAuth tokens, customer data, or provider keys.

python3 -m pytest -q

For web changes, see web/README.md.

Links

Dashboard · API · ClawHub · Umbrel app submission · License

About

Local-first AI model router for serious agents. One endpoint for OpenClaw, Codex, Claude Code, Cursor, Ollama, NVIDIA NIM, and authorized provider fallback.

Topics

Resources

Contributing

Security policy

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages