OmniRoute — The Free AI Gateway
One endpoint for 291 AI providers (90+ free), with automatic failover and built-in compression that saves 15–95% tokens — works with Claude Code, Codex, Cursor, and more.
- Source repo
- diegosouzapw/OmniRoute
- Stars
- ★ 78k
- Last updated
- today
- License
- MIT
- Primary language
- TypeScript
- FA score
- 60/100 · Some gaps
At a glance
- How it runs
- Works with
- Universal · cross-platformCodex · Claude Code · OpenAI API · Claude API
- Cost
- Free software; you pay for model usage
- Setup effort
- Low · running in minutes
- You'll need
- Typical use
- Developers who want a unified access point for multiple LLM providers, avoiding managing many SDKs and API keys.
- Not a fit if
- Users unwilling to send prompts to third-party AI providers
- Teams needing stable production-grade free quotas despite free-tier rate limits
- Source review
- 60/100 · Some gaps
What does this agent do, and when should you use it?
OmniRoute is an open-source, local-first AI gateway that aggregates 291 AI providers (90+ free, 40+ free forever) behind a single OpenAI-compatible endpoint (/v1). It connects 33 coding tools like Claude Code, Codex, Cursor, and Cline. With 19 routing strategies, including auto-combo and failover, it seamlessly switches providers when quotas run out. Built-in 12-layer pluggable compression (RTK, Caveman, etc.) averages ~89% token savings. It supports MCP (105 tools) and A2A protocols, and provides a CLI (80+ commands), dashboard, desktop app, and PWA. MIT-licensed, 100% local-first, and zero-config with free providers out of the box.
OmniRoute runs as a local proxy that receives requests from OpenAI-compatible clients and forwards them to configured providers through its smart router. It introduces 'Combos' (chains of models) with 19 strategies like priority, weighted, round-robin, cost-optimized, cache-optimized, etc. The default auto model scores connected providers live based on 12 factors and automatically falls back on failures. It compresses requests/responses via 12 engines (RTK, Caveman) to save tokens. It also offers an MCP server (stdio/HTTP/SSE) and A2A endpoint, allowing AI agents to manage routing, providers, combos, cache, and compression. The CLI provides 80+ commands for managing providers, keys, combos, logs, etc.
- Developers who want a unified access point for multiple LLM providers, avoiding managing many SDKs and API keys.
- Users of Claude Code or Codex who want automatic fallback to free or low-cost models when their quotas expire.
- Teams looking to optimize token consumption and reduce costs, saving up to 95% on token spend via compression.
- Scenarios requiring intelligent routing across providers based on latency, cost, or remaining quota.
- Developers seeking a local-first, privacy-preserving gateway for their coding assistants, without cloud dependencies.
How do you install or deploy this agent?
Install globally via npm: npm install -g omniroute. Then run omniroute to start the server on localhost:20128. Alternatively, use Docker: docker run -p 20128:20128 diegosouzapw/omniroute. Supports Windows, macOS, Linux, and even Android via Termux or as a desktop Electron app.
How do you use this agent?
After installation, point your tool to http://localhost:20128/v1. For zero-config, set your model to auto and OmniRoute will pick a provider automatically. Test with: curl http://localhost:20128/v1/chat/completions -H "Content-Type: application/json" -d '{"model":"auto","messages":[{"role":"user","content":"Hello!"}]}'. Configure supported tools with the base URL http://localhost:20128/v1. Use the CLI for management, e.g., omniroute combo list to view combos, or omniroute setup for guided setup.
What are this agent's strengths and limitations?
- Aggregates 291 providers with 90+ free tiers, zero-config start
- 19 routing strategies and automatic failover for high reliability
- 12-layer compression averaging ~89% token savings
- MCP and A2A support for complete agent control
- Local-first, privacy-focused, no cloud dependency
- MIT-licensed, fully open-source, self-hostable
- Free provider rate limits and quotas can lead to variability
- Initial configuration may be complex despite zero-config option
- Large feature set may have a steep learning curve
- Some advanced features (e.g., LLMLingua-2) may add latency
- Requires Node.js environment and network access to providers
How does this agent compare with similar options?
Compared to alternatives like OpenRouter and LiteLLM, OmniRoute offers a more comprehensive feature set with 291 providers, 19 routing strategies, built-in compression, and MCP/A2A support. It emphasizes local-first and free providers over OpenRouter's centralized approach, while LiteLLM is a lighter-weight library.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| OmniRoute — The Free AI Gateway This agent | 60 · Some gaps | CLIFree + model costs | ★ 78k | today | TypeScript | Codex · Claude Code · OpenAI API · Claude API |
| OpenSquilla — Token-Efficient AI Agent | 51 · Major gaps | CLIFree + model costs | ★ 7.1k | 1d ago | Python | OpenAI API · Claude API |
| ClawCodex | 51 · Major gaps | CLIFree + model costs | ★ 911 | today | TypeScript | ChatGPT · OpenAI API · Claude API |
| Code Puppy | 57 · Major gaps | CLIFree + model costs | ★ 869 | today | Python | Codex · OpenAI API · Claude API |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows security architecture (AES-256-GCM encryption, JWT, OAuth, guardrails), but least privilege principle not explicitly documented; user confirmation mechanisms (e.g., for dangerous actions) not clear; data flow transparency documented but not detailed; sensitive data handling has encryption and PII redaction, but default configuration not specified; dependency security has npm audit and vulnerability checks, but specific dependency versions not provided; external effects (e.g., MITM, TLS fingerprinting) mentioned but not fully explained; rollback mechanism not clear; source attribution has author and repository info, but publisher unverified.
Self-consistency: README and SECURITY.md descriptions align, but code implementation not verified; dependency availability: package.json and CI present, but dependency fetchability not verified; failure messages: error handling documented, but actual error messages not verified.
Audience and scenarios: README clearly targets developers, provides multiple use cases; capability boundaries: documentation describes feature scope, but limitations not explicit; trigger precision: routing strategies and model selection documented, but not verified; environment fit: multiple installation methods supported, but not verified.
Information architecture: README, docs directory, clear structure; install notes: Quick Start and multiple installation methods; naming stability: version number present, but API stability not verified; examples and FAQ: examples provided, but no FAQ; known limitations: not explicitly listed; license: MIT; versioning and changelog: version number present, but no changelog; maintenance responsibility: maintainer info present, but not verified.
Output usability: API and CLI present, but output format not verified; marginal value: free AI gateway with unique features; cost benefit: claims cost savings, but not verified.
Claim traceability: claims in README, but no evidence; cross-source corroboration: no multiple sources; fact/inference separation: documentation distinguishes, but not verified.
- Publisher identity unverified; proceed with caution.
- Security claims (encryption, guardrails) not verified against code; based on documentation only.
- Dependency security lacks specific versions; conduct your own audit.
- External effects (MITM, TLS fingerprinting) may raise legal/ethical concerns.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
Release notes excerpt
OmniRoute v3.8.51 2,022 documented changes — 235 features, 1,454 fixes and 333 maintenance entries from 1,973 commits in the cycle, with 301 external contributors. Thank you all. 📄 Complete release notes (every entry, full description and credit): attached below as CHANGELOG-v3.8.51.md, and in [CHANGELOG.md → [3.8.51]](https://github.com/diegosouzapw/OmniRoute/blob/main/CHANGELOG.md). GitHub caps a release body at 125,000 characters, so the 1,454 fixes and 333 maintenance entries live there in full. 📊 Release by the numbers | | | | --- | ---: | …View on GitHub ↗ -
Release notes excerpt
📊 Release by the numbers | | | | --- | ---: | | 👥 People who contributed | 248 | | 📝 Commits in the cycle | 1,714 | | 🔀 Pull requests referenced | 1,666 | | 📋 Changelog entries | 1,182 | | 🙌 Contributors credited in entries | 256 | | 🤖 Automated dependency commits | 22 | Entries by type | Type | Count | | --- | ---: | | 🐛 Fixes | 779 | | ✨ Features | 169 | | 📚 Docs | 29 | | 🧹 Chore | 27 | | 🧪 Tests | 15 | | ♻️ Refactor | 5 | | ⚡ Performance | 3 | | providers | 2 | | 🔒 Security | 2 | | ⚙️ CI | 2 | | deps | 2 | | maint | 2 | 🏆 Top 25 contributors this cycle …View on GitHub ↗ -
radar-export-latest Radar catalog export (rolling)
Release notes excerpt
Export estável do catálogo OmniRoute para o Radar. Atualizado automaticamente; NÃO é um release de versão do produto.View on GitHub ↗