All tools
Agentsby Diego Rodrigues de Sa e Souza (diegosouzapw) / GitHub
OmniRoute
Free
Free open-source AI gateway with smart auto-fallback across 237 providers and token compression
OmniRoute is a self-hosted, MIT-licensed AI gateway that unifies 237+ AI providers (90+ with free tiers) behind one OpenAI-compatible endpoint. It enables Claude Code, Codex, Cursor, Cline, Copilot, and 16+ other coding agents to route requests intelligently with automatic fallback when quotas are exhausted. Features RTK+Caveman stacked compression (15–95% token savings), 4-tier smart auto-fallback (Subscription → API → Cheap → Free), circuit breakers, MCP/A2A integration, and aggregates ~1.6 billion documented free tokens per month across provider pools.
Who it's for
Developers seeking cost-free AI codingHobbyist coders who can't justify $200+/month AI subscriptionsTeams wanting centralized AI provider managementMulti-model agent builders needing automatic failoverDevelopers building with Claude Code, Cursor, Copilot, or Cline
Pricing · free
checked today| Plan | Price | Includes |
|---|---|---|
| Open Source (Self-Hosted) | Free | MIT licensed · Fully free to run locally or on your own server · No credit card required · 237+ providers (90+ with free tiers) · ~1.6B free tokens/month aggregated · RTK+Caveman compression (15-95% token savings) · 4-tier smart auto-fallback · MCP integration (95 tools) · A2A agent protocol support · Desktop/PWA versions available · 17 routing strategies · Local-first privacy (data stays on your machine) |
AI-researched pricing — verify on the official site before subscribing.
Use it for
- — Routing Claude Code/Codex/Cursor/Cline requests to free tier providers when paid quotas are exhausted
- — Building resilient coding agents with zero downtime via provider auto-fallback
- — Reducing token consumption 15–95% through intelligent prompt compression
- — Aggregating and tracking 90+ provider free tiers from one dashboard
- — Splitting one subscription (e.g. Codex Pro) fairly across a development team
- — Eliminating rate-limit frustration in AI-assisted coding workflows
- — Bridging multiple AI providers (Claude, GPT, Gemini, DeepSeek) via single API endpoint
- — Integrating 95+ MCP tools for autonomous agent control of the gateway
Get the most out of it
- 01Start with auto mode (model: auto/coding or auto/fast) to let OmniRoute score providers live on 9 factors (health, quota, cost, latency, success rate); no manual combo assembly needed
- 02Build a 4-tier combo (Tier 1: your paid subscription quota, Tier 2: cheap backup API, Tier 3: budget provider like GLM, Tier 4: free forever provider) to achieve zero downtime when primary quotas exhaust
- 03Connect at least one free-forever provider (Kiro AI, OpenCode Zen, Qoder, Pollinations) with zero signup friction—use the Dashboard Providers tab to test immediately before configuring your tool
- 04Enable RTK+Caveman compression on tool-heavy sessions (git diffs, grep logs, file listings) where it achieves 78–95% token savings; watch the live /dashboard/compression stats to validate gains on your workload
- 05Set Base URL to http://localhost:20128/v1 in Claude Code/Cursor/Cline and copy your dashboard API key—OmniRoute exposes a single OpenAI-compatible /v1 endpoint so minimal tool reconfiguration is needed
In the news
Was this useful?