AIToolScan

OmniRoute Overview

OmniRoute is a free, open-source AI gateway that unifies 236 LLM providers behind a single OpenAI-compatible endpoint. It auto-fallbacks across providers in milliseconds, drains free-tier quotas, and compresses tokens so coding agents like Cursor, Claude Code, and Codex never stop building.

• 236 Providers, One Endpoint: Route OpenAI, Claude, Gemini, and Responses API traffic through localhost:20128/v1 with automatic tier-1/2/3 fallback when quotas run out.
• 17 Routing Strategies: Use auto, /coding, /cheap, cost-optimized, round-robin, context-relay, and custom combos to match your workflow.
• Token Compression: Stacked RTK + Caveman pipeline cuts 15–95% of eligible tokens on tool output and repeated context.
• Free Quota Draining: Tracks 90+ free providers (~1.6B tokens/month) with pool-deduped, honest usage counting — 11 free forever.
• Built-in MCP & A2A: Exposes the gateway as 95 MCP tools across 30 scopes and an A2A server with 6 agent skills over JSON-RPC 2.0.
• 3-Layer Resilience: Circuit breaker per provider, cooldown per connection, and lockout per model — failures stay isolated.
• Runs Anywhere: Install via npm, Docker, Electron desktop, PWA, Termux, or source — dashboard and API on port 20128.
• Production-Grade: Hybrid memory (FTS5 + Qdrant), guardrails, evals, TLS stealth, proxy support, and 21,000+ automated tests — MIT licensed.