AI router is a stdio MCP server that lets AI coding agents (Cursor, Claude Code, Claude Desktop, and other MCP clients) send prompts to Gemini and ChatGPT through real browser sessions — no API keys
No reviews yetBe the first to leave a review for AI Router
How does it handle rate limits from the browser sessions, and is there a risk of getting flagged or blocked by Google or OpenAI when you route heavy agent traffic through them?
Report
Maker
Hi Sevil, thanks for the question. It doesn't actually bypass rate limits — it detects and serializes them. Each provider gets one browser tab with a FIFO queue, so you won't spam parallel prompts on a single session. When Google or OpenAI caps you, the server just passes back a RATE_LIMITED status (there's no auto-retry built-in)
Report
Maker
@sevilgwkc Also, just a heads-up: we're currently working on adding Claude provider support next, so stay tuned!
Report
The fact that it piggybacks on real browser sessions to skip API keys is genuinely clever, especially for anyone juggling multiple AI tools and hitting rate limits daily. Nice execution.
Report
Maker
@baskapanad54699 Hey Adem! Thanks for the kind words. You hit the nail on the head 🙏, piggybacking on real sessions is great for personal/dev use to avoid API key hassle. That said, for heavy production-scale traffic, there's always a real risk of getting flagged since it operates through the web session, so official APIs with proper quotas are still the way to go for big workloads!
Report
finally a way to use chatgpt from claude code without juggling api keys. it just works through the browser, which is honestly kind of clever.
Report
Maker
Hi@fevzi734517 .Thank you so much! That is exactly our goal: to provide robust support for software development while also making it easy to compare models to see which one provides the best results.
We are currently working on expanding support for Claude and are actively developing integrations for DeepSeek 🔥.
How does it handle rate limits from the browser sessions, and is there a risk of getting flagged or blocked by Google or OpenAI when you route heavy agent traffic through them?
Hi Sevil, thanks for the question. It doesn't actually bypass rate limits — it detects and serializes them. Each provider gets one browser tab with a FIFO queue, so you won't spam parallel prompts on a single session. When Google or OpenAI caps you, the server just passes back a RATE_LIMITED status (there's no auto-retry built-in)
@sevilgwkc Also, just a heads-up: we're currently working on adding Claude provider support next, so stay tuned!
The fact that it piggybacks on real browser sessions to skip API keys is genuinely clever, especially for anyone juggling multiple AI tools and hitting rate limits daily. Nice execution.
@baskapanad54699 Hey Adem! Thanks for the kind words. You hit the nail on the head 🙏, piggybacking on real sessions is great for personal/dev use to avoid API key hassle. That said, for heavy production-scale traffic, there's always a real risk of getting flagged since it operates through the web session, so official APIs with proper quotas are still the way to go for big workloads!
finally a way to use chatgpt from claude code without juggling api keys. it just works through the browser, which is honestly kind of clever.
Hi@fevzi734517 .Thank you so much! That is exactly our goal: to provide robust support for software development while also making it easy to compare models to see which one provides the best results.
We are currently working on expanding support for Claude and are actively developing integrations for DeepSeek 🔥.