Prodigy

Scriptable, self-hosted data annotation tool with active learning

Hand-reviewed by the Toolglade team·Updated Aug 2026

data-labeling#data-annotation#nlp#spacy#active-learning

Claimed API Self-hosted Teams

Toolglade’s take

Prodigy is refreshingly different in a subscription-heavy market: a pay-once lifetime license and fully self-hosted, private data appeal strongly to NLP teams and researchers. Its scriptable recipes and active learning make it far more flexible than click-only labeling UIs, especially with spaCy. The trade-offs are a developer-oriented setup (it is a tool for people comfortable in Python) and its NLP-first heritage, though it does handle images and custom tasks. Best for teams that want to own their annotation workflow end to end.

About Prodigy

Prodigy is a scriptable, self-hosted annotation tool from Explosion for building ML training data, with active learning and tight spaCy integration, sold under a pay-once lifetime license.

Prodigy, built by Explosion (the makers of spaCy), is a developer-focused annotation tool designed to make creating training data fast enough that data scientists can do the labeling themselves. Its active-learning approach surfaces the most informative examples, enabling rapid iteration between annotating, training, and evaluating models. Prodigy is fully scriptable via recipes, so teams can customize labeling workflows for named entity recognition, text classification, part-of-speech tagging, dependency parsing, entity linking, coreference resolution, relation extraction, image classification, and custom HTML tasks. Its web app lets annotators work in the browser, including on mobile, while all data stays on your own servers. Distinctively, Prodigy is a downloadable developer tool with a pay-once lifetime license rather than a subscription, and it keeps all data private on your infrastructure. This makes it a strong fit for NLP-heavy and privacy-conscious teams that want a scriptable, self-hosted labeling workflow integrated with spaCy.

TL;DR

Prodigy is a scriptable, self-hosted annotation tool from the spaCy team, with active learning and a pay-once lifetime license, ideal for NLP teams valuing data privacy.

Company overview

Prodigy is built by Explosion, the company behind the popular open-source spaCy NLP library. It reflects Explosion's developer-first philosophy, providing a scriptable annotation tool that data scientists can run themselves.

Explosion positions Prodigy as a private, self-hosted alternative to cloud labeling platforms, with a lifetime license model rather than subscriptions.

Product features

Prodigy offers scriptable recipes, active learning, and a browser-based annotation UI supporting NER, classification, parsing, entity linking, coreference, relation extraction, and image tasks. It integrates tightly with spaCy for rapid model iteration.

All data stays on the user's own machines, and the tool is sold as a downloadable product with a one-time lifetime license.