BRAT

Input—per 1M tokens

Output—per 1M tokens

Context—tokens

WeightsClosed

About

BRAT is ranked #12 of 21 in AI data labeling tools on Inferse. It runs on Linux, macOS, Self-hosted, Web. There is a free plan.

Compared on AI data labeling tools

Supported modalities: textbrat.nlplab.org
Model-assisted labeling: Yesbrat.nlplab.org
Human review workflows: Yesbrat.nlplab.org
Custom ontologies: Yesbrat.nlplab.org
Deployment options: self hostedbrat.nlplab.org
API access: Yesbrat.nlplab.org

Facts

Purpose: brat is a web-based tool for structured text annotation, where annotations have fixed forms that computers can process.brat.nlplab.org · 2 Oct 2026
Annotation types: It supports text spans, binary relations, n-ary associations, attributes, normalization annotations, and freeform notes.brat.nlplab.org · 2 Oct 2026
Configurable: Annotation categories, types, and constraints are configurable for different text annotation tasks.brat.nlplab.org · 2 Oct 2026
Collaboration: Multiple annotators can work simultaneously on the same collection or document and see each other’s edits nearly as they are made.brat.nlplab.org · 2 Oct 2026
External resources: Normalization can associate annotations with external resources such as Freebase, Wikipedia, and Open Biomedical Ontologies.brat.nlplab.org · 2 Oct 2026
Automatic annotation: The tool can integrate output from automatic text annotation services accessible as web services.brat.nlplab.org · 2 Oct 2026
Languages: The server and client support Unicode, including nearly 100 different scripts.brat.nlplab.org · 2 Oct 2026
Exports: Annotations can be exported in brat stand-off format, and visualizations in SVG, PNG, PDF, or EPS.brat.nlplab.org · 2 Oct 2026
Validation and search: brat validates annotations against configured constraints after edits and supports searches across documents or collections, including key-word-in-context concordancing.brat.nlplab.org · 2 Oct 2026
Hosting requirements: A new server installation assumes a UNIX-like environment; the installation guide says Windows users can run it in a virtual machine with a UNIX-like operating system.brat.nlplab.org · 2 Oct 2026
Security limitation: The experimental standalone server is not security-hardened and is not recommended for sensitive data or systems accessible from the internet; the guide recommends a full web server for production.brat.nlplab.org · 2 Oct 2026
Support: The project directs usage questions to its mailing list and bug reports or feature requests to its GitHub issue tracker.brat.nlplab.org · 2 Oct 2026
Project origin: The project says it was developed collaboratively by researchers from several groups interested in text annotation.brat.nlplab.org · 2 Oct 2026