Prodigy
Explosion's scriptable annotation tool: active-learning-powered labeling for NLP and beyond.
About
Prodigy is a downloadable annotation tool from Explosion, the Berlin company behind spaCy. Its premise is machine teaching: instead of handing labeling to an outsourced team in a waterfall process, data scientists annotate themselves with a model in the loop that prioritizes the most uncertain examples via active learning. It covers NER, text classification, POS tagging, dependency parsing, entity linking, coreference, relation extraction, and image tasks, with first-class spaCy integration, Hugging Face plugins, and LLM-assisted workflows.
Prodigy is commercial software sold as a one-time fee (a 5-seat company license is listed at $2,450) with 12 months of upgrades, not a subscription. Everything runs on your own machines with no data leaving your network, so it works fully air-gapped. It assumes a developer user comfortable with Python and the command line, which is its main limitation for non-technical labeling teams.
What it can do
- Active-learning-driven annotation recipes
- NER, classification, parsing, coreference, and image tasks
- spaCy and Hugging Face model integration
- Fully local and air-gapped operation
- Python API and CLI for custom workflows
Setup
- Buy a license at prodi.gy (one-time fee, 12 months of upgrades).
- Install with pip using your download link.
- Start a recipe, e.g.
prodigy ner.manual dataset blank:en ./data.jsonl --label PERSON,ORG. - Annotate in the web UI, then train with
prodigy train.
Prerequisites
A Prodigy license (one-time fee); Python 3.8+.
Use cases
Works with
Trust score — 7/10
Scored 7/10 (verified tier, metrics-v1). Evidence: built by Explosion (spaCy makers), prodi.gy verified live; on sale since Dec 2017 per official FAQ; pricing public ($2,450 for 5 company seats). Strengths: battle-tested in NLP, active learning genuinely cuts labeling effort, air-gapped operation. Docked for: (1) commercial one-time license, no free tier; (2) assumes Python/CLI fluency, not for non-technical teams. License: commercial (one-time).
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 7.0/10
- Maintenance · recency, releases, issue load
- 7.0/10
- Openness · license + docs
- 4.0/10
- Manual review · sanity read + red-flag check
- 7.0/10
- · Built by Explosion (spaCy makers); prodi.gy verified live
- · On general sale since Dec 2017 (official FAQ)
- · Pricing public: $2,450 for 5 company seats (Sendowl)
- · Runs fully local/air-gapped; no data leaves the machine
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.