A single Rust binary that gives any AI SDK one endpoint to reach many LLM vendors, with MCP tools, traffic control and request logging.
598+170 this week94RustApache-2.0Well maintainedUpdated 23 hours ago
cheaper-inference/OmniRoute
Self-hostable MIT AI gateway that exposes one OpenAI-compatible endpoint to hundreds of model providers, with quota-aware auto-fallback and token compression.
Open-source alternative to:OpenRouter
- Stars
- 85,398
- Forks
- 12,901
- Open issues
- 540
- Last commit
- 5 hours ago
What is it?
OmniRoute is a self-hosted AI gateway that sits between coding tools and model providers. Tools such as Claude Code, Codex, Cursor, Cline and GitHub Copilot send requests to one local endpoint at localhost:20128/v1, and OmniRoute forwards them to a configured provider. It supports a broad list of providers and models, including Claude, GPT, Gemini, Kimi, GLM, DeepSeek and MiniMax.
When a provider hits a rate limit, runs out of quota or fails, OmniRoute can move the request to the next eligible target. Routing is organized into combos with 20 strategies, plus an auto mode that scores connected providers live. Built-in circuit breakers, per-connection cooldowns and per-model lockouts isolate failures. Optional RTK and Caveman compression reduce tool output and prompt size before requests are sent.
The project also includes a dashboard, an MCP server, A2A support, and desktop and PWA builds. Users bring their own provider accounts or API keys; availability of free tiers depends on each provider's terms.
Key features
- 01Single OpenAI-compatible endpoint in front of many model providers
- 02Automatic fallback across providers when quotas, rate limits or errors occur
- 0320 routing strategies configurable per combo, plus an auto mode
- 04Optional RTK and Caveman compression to reduce token usage
- 05Built-in MCP server and A2A protocol support
- 06Web dashboard, desktop app and PWA
Use cases
- Pointing Claude Code, Cursor or Cline at one local endpoint while switching between providers
- Keeping a coding session running when one provider hits its rate limit
- Sharing a set of provider keys across a small team with usage limits
- Comparing or routing requests across several models behind one API
Quick start
- 1
Install OmniRoute from the npm package named omniroute, or use the Docker Hub image, as described in the README
Things to know
- You must bring your own provider accounts and credentials, and each provider sets its own free-tier limits and terms.
- The headline token figures are estimates across third-party free tiers that can change or end at any time.
In collections
Topics
FAQ
- What is OmniRoute?
- Self-hostable MIT AI gateway that exposes one OpenAI-compatible endpoint to hundreds of model providers, with quota-aware auto-fallback and token compression. OmniRoute is a self-hosted AI gateway that sits between coding tools and model providers. Tools such as Claude Code, Codex, Cursor, Cline and GitHub Copilot send requests to one local endpoint at localhost:20128/v1, and OmniRoute forwards them to a configured provider. It supports a broad list of providers and models, including Claude, GPT, Gemini, Kimi, GLM, DeepSeek and MiniMax.
- Is OmniRoute free to use? What license does it have?
- OmniRoute is open source under the MIT license. Check the license terms before using it in commercial products.
- Is OmniRoute still maintained?
- Actively maintained. The last commit to OmniRoute was on Nov 7, 2026.
- What can I use OmniRoute for?
- Pointing Claude Code, Cursor or Cline at one local endpoint while switching between providers Keeping a coding session running when one provider hits its rate limit Sharing a set of provider keys across a small team with usage limits Comparing or routing requests across several models behind one API
- How do I install OmniRoute?
- Install OmniRoute from the npm package named omniroute, or use the Docker Hub image, as described in the README
- Where can I get OmniRoute?
- The source code and installation instructions are on GitHub: https://github.com/cheaper-inference/OmniRoute
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
Self-hosted, OpenAI-compatible LLM gateway that routes one endpoint to 40+ providers with virtual keys, budgets, and usage tracking.
865+83 this week113PythonApache-2.0Updated 7 hours ago
A self-hosted LLM API gateway for individuals that aggregates multiple provider channels behind one endpoint with failover and cost tracking.
2.7K+9 this week450TypeScriptAGPL-3.0Well maintainedUpdated 16 days ago
An open runtime that discovers, reports on, and routes the coding agents already running on a developer's machine, with per-agent cost tracking.
11.9K+588 this week3.3KRustOtherUpdated 16 days ago
Open-source AI gateway giving one OpenAI-compatible API across 23+ LLM providers, with failover, load balancing, semantic caching and virtual keys.
9.4K+173 this week1.5KGoApache-2.0Updated 11 hours ago
Open-source Go gateway that routes LLM requests across 30 providers through one OpenAI-compatible API, with caching, guardrails and cost controls.
280+6 this week40GoApache-2.0Well maintainedUpdated 5 days ago