Pick one tool per job. Get a recommended, shareable, agent-readable stack — with pricing, MCP, and whether an agent can sign up on its own. 248 tools, one decision at a time.
The framework that runs your agent's loop.
Where your model calls actually go.
Long-term memory across sessions.
Live web + document search for grounding.
Embeddings + similarity search.
See what the agent does + measure quality.
Run agent-generated code safely.
Let it drive a real browser.
Talk and listen.
Let the agent send + receive messages.
Let it act inside third-party apps.
Charge for what you built.
A fast way to assemble a working AI-agent stack: pick one tool per job (framework, LLM, memory, retrieval, vector store, evals, and optional pieces) and get a recommended, shareable build sheet with links, pricing, and whether an agent can sign up on its own.
Defaults are the tools most founders reach for first in each category; the alternatives are ranked by community traction (GitHub stars) and filtered by your preference (open source, hosted API, or agent-self-signup). Every pick links to its full data page.
Yes. Any stack is addressable as JSON at /api/stack.json (pass your picks as query params), so an agent can request a recommended stack and act on it programmatically.
Yes — the Explorer and every tool page are free. Individual tools have their own pricing, shown on each pick.
New how-tos, tutorials, and the tools worth your time — free, once a week. No spam, no scrape.