Autonomous discovery and classification engine for AI agent repositories.
PredAItor crawls GitHub, HuggingFace, and the web to find agentic systems, extracts their technical capabilities, filters by quality signals, and exports structured profiles for analysis.
Built to answer: what agents actually exist, what they can do, and which ones are worth paying attention to.
- Discovery — Finds agent repositories across GitHub, HuggingFace, and research paper sources
- Feature detection — Extracts technical capabilities: tool use, memory, autonomy level, model dependencies
- Quality filtering — Evaluates repository maturity and signal density
- Dataset export — Outputs standardized agent profiles to JSON for downstream analysis
- Node.js + TypeScript
- Cheerio for web scraping
- SQLite for local dataset storage
- Axios for HTTP
- Express API for integration
git clone https://github.com/AngelReml/PredAItor.git
cd PredAItor
npm install
cp .env.example .env
npm run devAgent profiles are exported to data/agents_dataset.json with
standardized fields: capabilities, maturity score, source, and
technical stack.
MIT