getagentictools

The catalog, from your terminal

agentictools is a zero-dependency CLI over all 41,734 listed tools — real installs and GitHub stars, refreshed weekly. No install needed, no telemetry, and a stable --json contract your agents can use too.

Try it

npx agentictools search pdf extraction
npx agentictools search browser -c mcp
npx agentictools info skills/mattpocock-skills-tdd
npx agentictools install <id>        # prints the command — never executes
npx agentictools stats

For agents & scripts

Every command takes --json; results go to stdout, status to stderr: 

npx agentictools search "sql database" -c mcp --json | jq -r '.[0].url'

Agent speaks MCP instead? The same catalog is a remote MCP server.

Links

Frequently asked questions

What is the agentictools CLI?

A zero-dependency command-line tool for searching the getagentictools catalog — 41,734 agent skills, MCP servers, Claude Code plugins and agentic loops — from your terminal. Run it without installing: npx agentictools search <query>.

Does it execute install commands?

No. "agentictools install <id>" prints the tool's install command and tells you where it runs (your shell, or inside a Claude Code session for slash commands) — it never executes anything itself. Half the catalog installs via Claude Code slash commands, so printing is the only honest behavior.

Can agents use it?

Yes — that is a design goal. Every command takes --json and prints results to stdout only (status goes to stderr), so output pipes cleanly. An agent that can run "npx agentictools search --json <query>" can discover tools on demand. If your agent speaks MCP, the same catalog is also available as a remote MCP server.

Does it phone home?

No telemetry, no accounts. The only network call is one JSON fetch of the catalog index from getagentictools.com, cached locally for 24 hours with ETag revalidation. Offline runs use the cache.