Developer first
Build your products on top.
Designed for developers to launch products on top.
text
AI text classifier,
for developers.
API first classifier built for developers. Classify human text with 99.98% accuracy. Fine-tune, integrate or build your own classifier toolset on top. Trusted by engineers from leading AI labs. Starts at $65/mTok.
Fine-tune models
Custom model fine-tuning
Our advanced classifier tunes up with a few human samples.
Classifier design
Token patterns, not perplexity.
Analyze token patterns, so odd grammar choices will be ok.
$65/1mTok.
API first, developer friendly pricing.
No logs, ever
We don't store evals.
No text is logged on our servers, whatsoever.
Data residency
Pick EU/US data processing.
Select a region that works for compliance.
Research Preview
Use closed domain tech.
Get access to top-tier lab tech at a fraction of the cost.
No text stored.
Choose between US/EU servers for data residency.
Validate snippets
Auto-sample words from large documents.
Use random sampling to evaluate content at scale.
Low latency
Verify millions of documents per minute.
Our developer API scales up to maintain low latency.
Reduce cost
Get the highest per token return on classification.
Why spend validating 1m words from the same website?
Sample from large batches.
Use our sampling API route to save cost on non-critical classification.
Answers in short
Can I build my own AI detector app on top?
Yes absolutely! you can build your own AI detector workflows on top of our base model or on top of some of our more powerful niche detection models.
Can you fine-tune a model for my use-case?
Yes, our unique approach means we can fine-tune models with less than 1,000 human samples - making us incredibly useful for a wide variety of niche cases.
Why is price per million tokens not words?
Different text have different word to token ratios, so to keep pricing consistent we use tokens as the base.
Why do you hate AI generated content?
We don't! We just think it's dangerous if AI content masquerades as Human written text without labels.
What languages do you support?
At the moment our base model only supports English dialects, but we do have specific models for other languages.
What is AI assisted? and why is it different?
It is human content edited by AI, it usually doesn't flag up on structural filters since the text is structurally human.