The npm registry
for AI agents.
Discover, evaluate, and integrate 500+ AI agents, Claude skills, and MCP tools — quality-scored by an LLM eval harness, personalized to your stack.
0+
agents indexed
< 0ms
p95 search latency
0
HuggingFace task types
0+
golden eval queries
Explore by HuggingFace task.
Text Generation
Produce free-form text from a prompt.
Sentence Similarity
Measure how close two pieces of text are in meaning.
Summarization
Condense long text into a shorter version.
Text Classification
Assign predefined labels to text.
Feature Extraction
Turn text into dense numeric vectors.
Image-Text-to-Text
Reason over images and text together.
Zero-Shot Classification
Classify into labels the model was never trained on.
Text Ranking
Reorder candidates by relevance to a query.
Everything the ecosystem
was missing.
Search in plain English.
Describe what you need. Agentify maps your query across 500+ indexed tools using hybrid BM25 + semantic retrieval. No keyword guessing. No hours wasted.
Scored, not just starred.
Every agent is evaluated against 50+ real-world queries. Scores reflect Precision@5, NDCG, and MRR — not GitHub star count or last commit date.
Ranked for your stack.
Tell Agentify your language, framework, and use case. Results re-rank based on what developers with your exact profile actually shipped to production.
Scrape. Evaluate. Surface.
Scrape
Multi-source ingestion across GitHub repos, Reddit threads, HuggingFace model cards, and developer blogs. Runs daily via Prefect pipeline — zero manual curation.
Evaluate
LangGraph 5-agent pipeline: Parser → Embedder → Scorer → Ranker → Indexer. Each agent evaluates quality against a 50-query golden eval set with measurable P@5, MRR, NDCG.
Surface
Hybrid BM25 + pgvector semantic retrieval. Results personalized by developer stack profile. p95 latency under 200ms — search-engine fast, not RAG-pipeline slow.
Find your next
agent in seconds.
Search the registry, compare quality scores, and copy install commands — no sign-up required.