flexrouter pools the free tiers of several AI providers behind one local address.
$ pip install flexrouterShort answer: a tool that lets apps run on free AI without hitting the limits.
Google, Groq, Mistral and a few other AI companies give developers free access to their models, but the limits are tight. Some of Google's newest models allow 20 requests a day. Groq's free tier stops at around 30 a minute. That sounds like plenty until you realise a modern AI app, like a coding assistant or anything that uses tools, can fire off a dozen requests for a single thing you ask it. On one free account, it stalls within minutes.
flexrouter pools them. It runs in the background on your computer, and every app sends its AI requests to it instead of to one company. It tracks how much free usage is left everywhere and sends each request to the best model that still has room. When one hits its limit, the next one takes over, so the app keeps working and the bill stays at zero.
A dashboard shows where every request went, what's left for the day, and what broke and why.
Limits as of September 2026.
Shown on first start, and again whenever no model works. Each step ticks itself.
The same pages as in the story, on the real thing.
Your app asks for a bucket like smart or fast, never a model name. Each bucket ranks its models by score or by speed, then picks at random among the ones within 20% of the best that can answer, so the load spreads out.
When a model is rate-limited, down or broken, the next one answers straight away, with no waiting between tries. A request pinned to one model reports the failure instead of switching, so you always know which model answered.
Add more than one key for a provider and they are used in turn. A key that hits its limit is skipped until it resets. Keys live in their own file, never in your settings file.
Every model and key has one status: Ready, Busy, Struggling, Needs you or Off. Errors are explained in plain words. The error brain sorts each new error, says how sure it is, and asks you to check when it isn't.
See what is left of every free allowance today and when each limit resets. Limits you set are marked as yours. Numbers the provider reports carry the provider's name.
Pick a provider from the list, marked free or paid, and paste a key. Add models with AI finds the models the key can use, Find rate limits with AI fills in their limits, and Rank models with AI puts them in order.