WebRouter
More Than a Proxy. Your AI Backend.
5 followers
More Than a Proxy. Your AI Backend.
5 followers
Most "AI gateways" just forward requests. WebRouter is different — a full-stack AI backend that unifies 15+ LLM providers, remembers conversations, comes with built-in RAG knowledge base, team management with quotas, cost tracking, and health monitoring. Self-hosted with one Docker command. Open source (BSL-1.1).
Free
Launch Team
Framer AI AgentsDesign and publish professional sites with AI
Promoted
## Q: How is this different from a regular API proxy / relay?
Most API proxies just forward requests — they're transport-layer tools. WebRouter also handles **application-layer** needs: session memory so your AI remembers conversations, a built-in knowledge base (RAG) so your agents can look up your docs, multi-user team management with per-user quotas, real-time cost tracking, and health monitoring with auto-fallback.
Think of it as the difference between a universal remote (proxies) and a full home automation system (WebRouter).
---
## Q: Can I just use OpenAI directly? Why do I need this?
If you're a solo dev calling one model for a side project — you probably don't. But as soon as you have multiple providers, a team, or a production service, the friction adds up fast:
- Switching models means code changes without WebRouter
- Every provider has its own auth, rate limits, and pricing quirks
- No centralized cost visibility
- No built-in memory or knowledge base
- No fallback when a provider goes down
WebRouter is optional — until you hit any of these walls.
---
## Q: Is it OpenAI-compatible?
Yes. If your client speaks the OpenAI API format, it works with WebRouter. Just change the base URL:
```
https://api.openai.com/v1 → http://your-server:5051/v1
```
No SDK changes, no code changes, no prompt changes. Works with Claude Code, Cline, OpenClaw, LangChain, LlamaIndex, and any OpenAI-compatible client.
---
## Q: Can I use it with Claude Code / Cline / OpenClaw?
Yes. That's actually one of the primary use cases. Point your tool's API endpoint to WebRouter, and you can switch between providers (OpenAI, Anthropic, Google, DeepSeek, etc.) from the WebRouter admin panel — no tool config changes needed.
---
## Q: Do I need to set up a separate vector database for RAG?
No. WebRouter has a built-in knowledge base with semantic search. Upload your documents or point it to a directory — it handles chunking, vectorization, and indexing automatically. The retrieved context is injected into every request without you wiring up a separate vector DB pipeline.
---
## Q: Where does my data go?
Nowhere. WebRouter is self-hosted — you deploy it on your own server. All data (conversation history, knowledge base documents, logs, billing data) stays on your infrastructure. No third party sees your requests or your data.
Public relay services route your data through their servers. WebRouter doesn't.
---
## Q: What providers are supported?
15+ providers including: OpenAI (GPT-4o, GPT-4, GPT-3.5), Anthropic (Claude 3.5 Sonnet, Claude 3 Opus, Claude 3 Haiku), Google (Gemini 1.5 Pro, Gemini 1.5 Flash), DeepSeek (V2, R1), Alibaba Qwen (Max, Plus, Turbo), Azure OpenAI, AWS Bedrock, Together AI, Groq, and any OpenAI-compatible endpoint.
---
## Q: I don't want to self-host. Is there a cloud version?
Not yet. WebRouter is designed for self-hosting — one Docker command and you're up. A managed cloud version is on the roadmap but not available today.
---
## Q: What's the BSL 1.1 license? Is it really open source?
BSL 1.1 (Business Source License) means:
- You can freely use, modify, and deploy it in production — no cost
- You can't resell it as a SaaS offering
- After 4 years, it converts to Apache 2.0 automatically
It's "open source" in practice — the source is public, you can fork it, modify it, use it in production. The only restriction is on cloud providers packaging and reselling it. This protects the project from the "AWS wraps it and kills it" problem while being fully free for actual developers.
The integration module (wr-proxy) is MIT licensed, which is fully permissive.
---
## Q: How hard is it to set up?
Two commands:
```bash
git clone https://github.com/your-repo/web...
docker compose -f deploy/docker-compose.yml up -d
```
Then open `http://your-server:5050`, add your provider keys, and you're done. Takes about 5 minutes.
---
## Q: Can I use it behind a corporate firewall / air-gapped environment?
Yes. Since it's self-hosted and doesn't phone home, it works perfectly in air-gapped or restricted network environments. You only need outbound access to the AI provider APIs you configure.
---
## Q: Can I set usage limits per user / per team?
Yes. WebRouter has built-in multi-user support with three roles (admin, operator, read-only) and per-user / per-team monthly quotas. When a quota is hit, the API stops accepting requests — no surprise overruns.
---
## Q: What if a provider goes down?
WebRouter's health checker automatically detects provider failures and marks them as unavailable. With smart routing enabled, traffic is automatically redirected to healthy fallback providers. You can configure routing by cost priority, latency priority, or explicit fallback order.
---
## Q: How do I track costs?
Real-time. Every API call is metered and aggregated on the billing dashboard. You can view costs by model, by user, and by team — with daily, monthly, and trend views.
---
## Q: Is there a demo I can try?
Yes: **[demo.webrouter.tech](https://demo.webrouter.tech)** — login with `admin / admin123456`. It's a live instance so you can click around, see the dashboard, and test the provider management.
---
## Q: Can I contribute?
Yes. See CONTRIBUTING.md for guidelines. PRs are welcome. We also need help with documentation, translations, and integrations.