OmniRoute: The Free AI Gateway That Forked 9Router Into 290+ Providers
OmniRoute is an open-source AI gateway — one OpenAI-compatible endpoint that auto-falls back across 290+ LLM providers, with a compression pipeline and a desktop/PWA dashboard. It started as a TypeScript fork of 9Router.
- ⭐ 33037
- TypeScript
- Node.js
- Next.js
- React
- Electron
- MIT
- Updated 2026-07-29
9Router: Smart LLM Proxy with Token Saver • Orca: The ADE for Running Claude Code, Codex, and Cursor in Parallel

What Is OmniRoute? #
OmniRoute is an open-source AI gateway: install it once, point any OpenAI-compatible tool at its local endpoint (http://localhost:20128/v1), and it handles the mess of juggling multiple LLM providers for you — auto-falling back to another provider the instant one hits a rate limit, runs out of quota, or its key dies.
🔗 GitHub: https://github.com/diegosouzapw/OmniRoute 🌐 Homepage: https://omniroute.online
Note up front: OmniRoute began life as a TypeScript fork of 9Router (rtk-ai/rtk), a project
dibi8 already covered. The two have since diverged in different directions — OmniRoute added a much larger provider catalog, multi-modal APIs, and a polished desktop/PWA dashboard, while 9Router/RTK kept growing on its own and, as of late July 2026, actually has more stars (73,600+) than OmniRoute (33,000+). Worth knowing before assuming one is simply an upgrade of the other.
Zero Config on First Install #
A fresh install answers requests immediately, no API keys required, using keyless free backends wired into an auto routing mode:
npm i -g omniroute
curl http://localhost:20128/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{"model":"auto","messages":[{"role":"user","content":"Hello!"}]}'
How the Fallback Works #
OmniRoute’s routing sits in front of your provider accounts as a 4-tier cascade — it tries the cheapest/most-available tier first and falls back only when it has to:
| Tier | Source | Falls back when |
|---|---|---|
| 1. Subscription | Claude Code, Codex, Copilot plans you already pay for | Quota exhausted |
| 2. API Key | DeepSeek, Groq, xAI, etc. | Budget threshold hit |
| 3. Cheap | GLM ( | Budget threshold hit |
| 4. Free | Kiro, Qoder, Pollinations, and other free-tier backends | Always available as last resort |
Key Features #
| Feature | Description |
|---|---|
| 290+ Providers | One endpoint fans out across a large provider catalog, per the project’s own provider list (self-reported, not independently verified) |
| Compression Pipeline | Up to 12 engines (RTK, Caveman, LLMLingua-2, GCF, OmniGlyph) can compress prompts/tool output before they reach the model |
| MCP + A2A | Model Context Protocol (stdio/HTTP/SSE) and Agent2Agent v0.3 (JSON-RPC 2.0 + SSE) support |
| Local-first storage | SQLite (better-sqlite3, WAL) with AES-256-GCM encryption for stored keys |
| TLS stealth routing | JA3/JA4 TLS fingerprint impersonation via wreq-js, for reaching providers through restrictive networks |
| Multiple runtimes | CLI/server (Node.js), Electron desktop app, Android via Termux, and a browser PWA |
| Live analytics | Dashboard tracking usage, quota, savings, and p95 latency per provider |

Tech Stack (as documented by the project) #
- Runtime: Node.js 22.x/24.x LTS
- Language: TypeScript (project states 100% TypeScript in core, zero
anysince v2.0) - Framework: Next.js 16 + React 19 + Tailwind CSS 4
- Database: better-sqlite3 (WAL) + LowDB for legacy JSON, across 95 domain modules
- Auth: OAuth 2.0 (PKCE), JWT, API keys, MCP scoped auth
- Testing: Node.js test runner + Vitest — the project states 25,000+ test cases across 3,300+ files
Install Options #
npm i -g omniroute
Docker is also supported:
docker pull diegosouzapw/omniroute
Desktop (Electron), Android (Termux), and PWA builds are documented on the project’s GitHub releases and website.
Compatible Tools #
Any OpenAI-compatible client works by pointing at OmniRoute’s local endpoint. The project explicitly documents setup for 30+ tools, including:
Claude Code · Codex CLI · Cursor CLI · GitHub Copilot CLI · Cline · Kilo Code · Roo Code · Continue · Aider · OpenCode · Factory Droid · Goose · Hermes Agent · Grok Build

OmniRoute vs. 9Router #
| Aspect | OmniRoute | 9Router (RTK) |
|---|---|---|
| Relationship | Fork of 9Router | Original project |
| GitHub stars (late July 2026) | 33,000+ | 73,600+ |
| UI | Electron desktop + PWA dashboard | See dibi8’s 9Router article |
| Compression engines | RTK + Caveman + LLMLingua-2 + GCF + OmniGlyph | RTK (originating engine) |
| Protocols | MCP + A2A | See 9Router’s own docs |
| License | MIT | MIT |
Use Cases #
1. Never Losing Work to a Rate Limit #
Keep coding when Claude Code or Codex hits its quota mid-session — the 4-tier fallback picks up automatically instead of the request failing.
2. Trying Free Tiers Without Juggling Dashboards #
Instead of manually tracking a dozen providers’ free quotas, point one endpoint at OmniRoute and let it route to whichever free backend still has budget.
3. Reducing Token Spend on Tool-Heavy Agent Sessions #
The compression pipeline sits between your agent and the model, aiming to cut the token volume of large tool outputs before they’re billed.
Related Repositories #
| Repository | Purpose |
|---|---|
| 9Router (RTK) | The original project OmniRoute forked from — now the larger of the two by stars |
| Orca | Different problem (parallel agent orchestration) but a similar “one control point for many AI tools” philosophy |
Related Articles #
- 9Router: Smart LLM Proxy with Token Saver — the project OmniRoute forked from, now the bigger of the two
- Orca: The ADE for Running Claude Code, Codex, and Cursor in Parallel — a different angle on “one control layer for many AI tools”
Conclusion #
OmniRoute takes the “one endpoint, many providers” idea it inherited from 9Router and adds a much larger provider catalog, a compression pipeline, MCP/A2A support, and a polished Electron/PWA dashboard on top. It’s grown fast — from roughly 17,700 to 33,000+ stars in about two weeks as of late July 2026 — even though its parent project RTK remains the larger of the two. Worth trying if you want a dashboard-driven gateway rather than 9Router’s leaner core.
Best for: Developers who want a visual dashboard and broad provider coverage on top of the fallback-routing idea, and don’t mind a heavier, more actively-marketed project.
GitHub: https://github.com/diegosouzapw/OmniRoute
Recommended Infrastructure for Self-Hosting #
If you’d rather run OmniRoute’s server component on an always-on box instead of your laptop:
- DigitalOcean — $200 free credit for 60 days across 14+ global regions. A common choice for a persistent gateway endpoint.
- HTStack — Hong Kong VPS with low-latency access from mainland China. This is the same IDC that hosts dibi8.com — battle-tested in production.
Affiliate links — they don’t cost you extra and they help keep dibi8.com running.
Last updated: 2026-07-29
💬 Discussion