finds.dev

cheaper-inference/OmniRoute

Never stop coding. Free MIT AI gateway: one endpoint, 375 providers (150+ free), 1,400+ models Kimi, Claude, GPT, Gemini, GLM, DeepSeek, MiniMax. Works with Claude Code, Codex, Cursor, OpenCode, Cline & Copilot. Quota-aware auto-fallback, RTK+Caveman compression saves 15-95% tokens, MCP/A2A, Desktop/PWA. Built by 900+ contributors

View on GitHubHomepage ↗

★ 83,255TypeScriptMITupdated Nov 2026

// the take

OmniRoute is a self-hosted AI gateway that exposes one OpenAI/Anthropic-compatible endpoint and routes requests across hundreds of LLM providers, including stacked free-tier accounts, with automatic failover and token compression. It's aimed at developers who want to avoid vendor lock-in for coding assistants like Claude Code, Cursor, or Copilot, or who want to squeeze more mileage out of multiple free-tier API quotas.

The three-layer resilience design (provider circuit breaker, per-key cooldown with exponential backoff, per-model lockout) is a concrete engineering decision, not just marketing copy — a failing API key doesn't take down a whole provider connection, and a failing provider doesn't take down the combo. It solves a real annoyance: one OpenAI-compatible `/v1` endpoint instead of a dozen SDKs and rate limits for CLI coding tools. It's MIT-licensed and self-hostable via Docker/npm/source with no mandatory cloud dependency for the core routing path.

"TLS fingerprint stealth" and bypassing provider-side blocking of proxy clients is advertised as a core feature, not an edge case — that's explicitly about evading rate-limit/ToS enforcement by the AI providers themselves, and using it risks getting your own provider accounts banned. The headline "~1.62B free tokens/month" depends on stacking dozens of third-party free tiers that providers can revoke or throttle unilaterally; the README itself admits the number "moves both ways," so it's not capacity OmniRoute controls or guarantees. Running this centralizes API keys for every connected provider in one self-hosted process, so a compromise of this app compromises all of them — auditing a 380-provider, 1,400+ model attack surface is a lot to ask of any team, let alone a solo self-hoster. The README itself is built almost entirely around growth mechanics (star asks, affiliate signup links, 67 translated READMEs, donation addresses) rather than technical documentation, making it hard to judge real-world reliability independently of numbers the project computes and publishes about itself.

We read the README, the directory tree and a key source file in November 2026. Nobody pays to be listed. How we judge →

// at a glance

Stars
83,255
Forks
12,416
Language
TypeScript
License
MIT
Last commit
1 Nov 2026

// nearby

If OmniRoute isn't quite it.

All repos →