About catalog
catalog is a remote MCP server for searching agent skills, MCP servers, plugins, and agentic loops. It provides ranked, deterministic results with install commands and catalog links, using a dataset refreshed weekly.
Key features
- 01Search four tool categories
- 02View tool details and install commands
- 03Browse categories and live counts
- 04See recently updated entries
- 05Review ecosystem statistics
Use cases
- 01Find an MCP server for a specific service
- 02Discover skills for a task such as PDF extraction
MCP support
Exposes read-only catalog search and discovery tools to MCP clients over Streamable HTTP. The public endpoint requires no authentication.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://getagentictools.com/api/mcpclaude mcp add --transport http com-getagentictools-catalog https://getagentictools.com/api/mcp{"mcpServers":{"com-getagentictools-catalog":{"type":"http","url":"https://getagentictools.com/api/mcp"}}}Open in Cursor{"servers":{"com-getagentictools-catalog":{"type":"http","url":"https://getagentictools.com/api/mcp"}}}Open in VS Code{"mcpServers":{"com-getagentictools-catalog":{"type":"http","url":"https://getagentictools.com/api/mcp"}}}{"mcpServers":{"com-getagentictools-catalog":{"serverUrl":"https://getagentictools.com/api/mcp"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
Tools
5| Parameter | Type | Required | Description |
|---|---|---|---|
| limit | integer | No | Maximum results to return (1-25, default 10) |
| query | string | Yes | What to look for, e.g. "pdf extraction" or "postgres" |
| category | enum | No | Restrict results to one category; omit to search all fourskills | mcp | plugins | loops |
Details
From the project READMEgetagentictools MCP server
Search 41,000+ agent skills, MCP servers, Claude Code plugins, and agentic loops from inside any MCP-capable agent. This is the remote MCP server for getagentictools.com — real install counts and GitHub stars, refreshed weekly, never faked.
Endpoint: https://getagentictools.com/api/mcp — Streamable HTTP, free, no auth, no tracking.
Connect your client
Claude Code
claude mcp add --transport http getagentictools https://getagentictools.com/api/mcp
Cursor
// .cursor/mcp.json
{
"mcpServers": {
"getagentictools": { "url": "https://getagentictools.com/api/mcp" }
}
}
VS Code (Copilot)
// .vscode/mcp.json
{
"servers": {
"getagentictools": { "type": "http", "url": "https://getagentictools.com/api/mcp" }
}
}
OpenAI Codex
# ~/.codex/config.toml
[mcp_servers.getagentictools]
url = "https://getagentictools.com/api/mcp"
Gemini CLI
// ~/.gemini/settings.json
{
"mcpServers": {
"getagentictools": { "httpUrl": "https://getagentictools.com/api/mcp" }
}
}
Then ask your agent things like “find me an MCP server for Postgres” or “what skills exist for PDF extraction?” — results carry install commands and links to their catalog pages.
Tools
| Tool | What it does |
|---|---|
search_tools | Ranked search across all four categories. Query, optional category filter; install command and page URL on every hit |
get_tool | Full detail for one tool — description, repo, license, install command, plus SKILL.md-extracted capabilities for enriched skills |
list_categories | The four categories with live counts |
whats_new | Recently updated tools, by real upstream push date |
ecosystem_stats | Measured aggregates: maintainer count, license coverage, star concentration |
How it works
- Deterministic. No LLM sits between your agent and the data — tokenized field-weighted scoring, popularity as a tiebreak only.
- Fresh by construction. The server reads the same dataset the website serves, generated in the same deploy, so it cannot drift from what you see on getagentictools.com. The catalog is re-harvested weekly from GitHub and public registries.
- Free and unauthenticated. Public data, read-only, no accounts, no tracking beyond standard web logs.
About this repo
This repo mirrors the deployed server source (api/), which lives in a private monorepo alongside the site and data pipeline. PRs are welcome — we port accepted changes upstream; direct pushes here are overwritten by the next sync.
The data endpoints it reads (/data/agent/*.json) are public — you can build your own tooling against them. Attribution appreciated.
License
MIT © Daniel Moka
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.