# About classifier.dev
What classifier.dev is, why it exists, what it runs on, and who runs it.
classifier.dev is a zero-shot text classification service: you send texts and a list of labels over plain HTTP, and get back the label that fits each text and how sure the model is. Free use needs no API key or account, so the first call works the moment you read the example.
## Why it exists
Language models can classify anything they can see, but seeing is the expensive part. An agent with forty search results, a thousand log lines or a backlog of tickets has to pull all of it into context to judge it. This service classifies ten thousand things without reading them: one request, a second later, labels and confidences for all of them, and only the survivors get read.
The confidence is calibrated, so an agent can act on the sure answers and hand the rest to a person or a stronger model. Without that split you would have to trust every answer equally.
## What it runs on
The model is Jev, TypeSafe's decision model. It returns a calibrated probability per label instead of generating text. The smart tier re-asks the answers Jev was unsure about of a reasoning model. Both are measured on public test sets at https://classifier.dev/benchmark, including how the service compares with calling Jev directly.
It runs as a single Cloudflare Worker with no runtime dependencies. The worker, the CLI, the eval harness and these pages are open source at https://github.com/mrmps/classifier-dev.
## Who
Built and run by Michael Ryaboy (https://x.com/michael_chomsky), an independent developer in San Francisco. It started as a tool for agents that needed to filter search results before reading them, and got a public API once other people wanted the same thing.
## Contact and policies
https://classifier.dev/contact for how to reach a person, https://classifier.dev/privacy for what is and is not logged, https://classifier.dev/terms for what you agree to by using it, https://classifier.dev/pricing for what it costs (nothing) and the limits.