Skip to main content
Glama
plugmcp

plugmcp

Official
by plugmcp-ai

Ask your agent

Once it is connected, ask in plain words:

  • "Find an MCP server that lets Cursor query my Postgres database, and install it."

  • "Compare the GitHub and GitLab MCP servers."

  • "What MCP servers came out this week?"

  • "Give me the Claude Code command for the Linear MCP server."

  • "Find an agent skill for editing PDFs and give me the message to install it in Grok Bot."

Related MCP server: remoteopenclaw-mcp

Install

The fastest path needs no install at all. The directory is a remote MCP server at:

https://plugmcp.ai/mcp-server

The Cursor Marketplace plugin is .cursor-plugin/plugin.json. Its mcp.json points at that same URL. No key is stored in this repository. The server answers without one. A key is optional and lives only in the client that connects. See API key.

Use this package when your client only runs local (stdio) servers, or when you want the server in a lockfile.

Claude Code

claude mcp add --transport http plugmcp https://plugmcp.ai/mcp-server
# or local:
claude mcp add plugmcp -- npx -y plugmcp-mcp

Cursor, Windsurf, Claude Desktop, and any mcpServers JSON

{
  "mcpServers": {
    "plugmcp": {
      "command": "npx",
      "args": ["-y", "plugmcp-mcp"]
    }
  }
}

VS Code (.vscode/mcp.json)

{
  "servers": {
    "plugmcp": { "type": "http", "url": "https://plugmcp.ai/mcp-server" }
  }
}

Codex CLI

codex mcp add plugmcp --url https://plugmcp.ai/mcp-server

Grok Bot

Grok Bot has no config file. Send this in a Grok Bot chat and confirm the card:

Add a custom MCP server called plugmcp at https://plugmcp.ai/mcp-server

Every Bot on your account gets it; attach it to a task with @. Then ask for any server or skill with grok-bot as the client, and the reply is the message that adds that one too.

Grok Build and grok.com

  • Grok Build: grok mcp add --transport http plugmcp https://plugmcp.ai/mcp-server

  • grok.com: open Connectors, choose New Connector, then Custom, and paste https://plugmcp.ai/mcp-server.

ChatGPT

Settings, then Connectors, then Add. Name plugmcp, URL https://plugmcp.ai/mcp-server.

API key (optional)

Every tier is paced, so one busy agent cannot use the whole day at once:

Per minute

Per hour

Per day

No key (per IP)

5

10

20

Free account key

15

60

400

Paid plan key

300

no cap

20,000

When a limit is hit, the tool result says which one and when it resets. Calls without a key may also wait while the directory is busy; key holders keep working. Create a pb_live_ key at plugmcp.ai/app and pass it as PLUGMCP_API_KEY:

{
  "mcpServers": {
    "plugmcp": {
      "command": "npx",
      "args": ["-y", "plugmcp-mcp"],
      "env": { "PLUGMCP_API_KEY": "pb_live_..." }
    }
  }
}

For the remote URL, send it as Authorization: Bearer pb_live_.... Details are in auth.md.

Tools

Every tool is read-only.

Tool

What it does

search_servers

Keyword search with filters for remote, official, and client

recommend_servers

Ranked picks for a task described in plain words

get_server

One server: tools, auth, packages, health, and optionally the README

get_install

The exact config for Cursor, Claude Code, Claude Desktop, VS Code, Windsurf, ChatGPT, Codex, Gemini CLI, Grok Build, Grok Bot, or plain JSON

compare_servers

Side by side for two to five servers

whats_new

Servers added recently

list_categories

Categories with counts

search_skills

Search agent skills (SKILL.md)

get_skill

One skill with install commands for Cursor, Claude Code, Codex, and Grok Build, plus the Grok Bot message

How it works

plugmcp-mcp is a small stdio server built on the official MCP TypeScript SDK. It forwards tools/list and tools/call to https://plugmcp.ai/mcp-server over Streamable HTTP. The upstream connection opens on the first call and reconnects after a network error. Nothing is stored locally.

Variable

Default

Purpose

PLUGMCP_API_KEY

none

Optional pb_live_ key

PLUGMCP_URL

https://plugmcp.ai/mcp-server

Directory endpoint

Develop

npm ci
npm test        # build, then unit tests against an in-memory upstream
npm run smoke   # run the built server over stdio against plugmcp.ai

Release

Push a v* tag that matches package.json. The publish workflow runs the tests, publishes to npm (NPM_TOKEN secret), then publishes server.json to the MCP Registry as ai.plugmcp/directory, signed with the plugmcp.ai domain key (MCP_REGISTRY_PRIVATE_KEY secret).

License

MIT

Available Tools

9 tools
compare_serversCompare MCP serversA
Read-onlyIdempotent
Inspect

Compare two to five servers side by side by slug: trust signals, auth, hosting, transport, tools count, license.

Inputs: slugs (required), response_format. Returns: missing, servers

get_installGet install configA
Read-onlyIdempotent
Inspect

The exact config or command to install one server in one client: Cursor, Claude Code, Claude Desktop, VS Code, Windsurf, ChatGPT, Codex, Gemini CLI, Grok Build, Grok Bot, or plain JSON. For Grok Bot the value is the message to send in its chat. Includes the config file path and a one-click deeplink when the client has one.

Inputs: slug (required), client. Returns: auth, file, kind, name, +8 more

get_serverGet one MCP serverA
Read-onlyIdempotent
Inspect

Details for one server by slug: description, tools, auth, transport, health, packages, and links. include_readme adds the README (trimmed).

Inputs: slug (required), include_readme, response_format. Returns: url, name, slug, tools, +22 more

get_skillGet one agent skillA
Read-onlyIdempotent
Inspect

One skill by slug: what it does, where the SKILL.md lives, and the install command for Cursor, Claude Code, Codex and Grok Build, plus the chat message that installs it in Grok Bot. include_body adds the SKILL.md text (trimmed).

Inputs: slug (required), include_body, response_format. Returns: url, body, file, name, +7 more

list_categoriesList categoriesA
Read-onlyIdempotent
Inspect

Directory categories with listing counts. Use a slug as the category filter in search_servers.

Inputs: no named parameters. Returns: categories

recommend_serversRecommend MCP servers for a taskA
Read-onlyIdempotent
Inspect

Describe what the user wants to do in plain words. Returns the best three servers, each with why it fits, auth, remote or local, health, and the install snippet for the client.

Inputs: task (required), limit, client, constraints, +1 more. Returns: hint, task, servers, keywords

search_serversSearch MCP serversA
Read-onlyIdempotent
Inspect

Find MCP servers by product name, publisher, or topic. Ranked by relevance (exact name, then publisher, then tags and description); the product's own server comes first when there is one (jira finds Atlassian's), duplicates merged. Returns compact cards with trust signals; call get_install next. When the user describes a job rather than a product, recommend_servers is one call instead of two.

Inputs: limit, query, client, cursor, +7 more. Returns: hint, count, query, total, +3 more

search_skillsSearch agent skillsA
Read-onlyIdempotent
Inspect

Find agent skills (SKILL.md packs for Claude Code, Cursor, Codex, Grok Build and Grok Bot) by name, publisher or topic. Call get_skill for the install command.

Inputs: limit, query, cursor, category, +1 more. Returns: hint, count, query, total, +3 more

whats_newNew MCP serversA
Read-onlyIdempotent
Inspect

Servers added to the directory recently, best first, plus health changes. Defaults to the last 7 days.

Inputs: limit, since, cursor, category, +2 more. Returns: hint, count, since, total, +4 more

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.9/5.0

    Scored across 9 tools

    Disambiguation4/5

    Most tools have clearly distinct purposes: server discovery, server details, installation, categories, recent additions, comparison, recommendations, skill search, and skill details. The only mild overlap is search_servers versus recommend_servers, but the descriptions explicitly distinguish keyword/product search from plain-language job recommendations.

    Naming Consistency4/5

    The set mostly follows a readable snake_case verb_noun or verb_noun_plural pattern: search_servers, get_server, compare_servers, recommend_servers, search_skills, get_skill. Minor deviations are whats_new as a phrase-like name and get_install being slightly less parallel, but the convention is still predictable.

    Tool Count5/5

    Nine tools is well-scoped for a directory covering MCP servers and agent skills. Each tool covers a meaningful operation, and the set avoids both a thin surface and an excessive number of near-duplicate tools.

    Completeness4/5

    The surface covers the main consumer workflows: search, inspect, install, categorize, discover new items, compare, recommend, and handle skills. Minor gaps exist, such as no dedicated skill comparison or broader browsing tool beyond search/categories/whats_new, but these are workable given the available filters and detail tools.

    Maintenance

    ActivityMaintained
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      A
      quality
      D
      maintenance
      Enables AI agents to search a directory of 13,870+ MCP servers, 4,384+ agent skills, and plugins from Remote OpenClaw, returning install commands directly.
      3
      67 npm
      7
      MIT
    • A
      license
      Not graded
      quality
      C
      maintenance
      Enables searching and retrieving details of 41,000+ agent skills, MCP servers, Claude Code plugins, and agentic loops from any MCP-capable agent.
      MIT
    • A
      license
      A
      quality
      C
      maintenance
      Enables coding agents to search an index of 22k+ MCP servers, agent skills and coding-agent plugins directly from chat, returning matches with stars, install commands and page links, plus trending repos and curated best-of topic lists. Read-only, no API key required, usable via a hosted HTTP endpoint or locally over stdio.
      4
      MIT