Skip to main content
Glama

curatedmcp

npm version npm downloads License: MIT Node.js β‰₯18

The CuratedMCP Agent. One CLI to discover, run, audit, and govern every MCP server your AI tools (Claude, Cursor, Windsurf, Copilot, Gemini) use.

# 10-second risk scan of your machine β€” no signup
npx curatedmcp audit

Plug it in once. Add servers anytime. Audit and govern them from one place.


What you get

Command

What it does

curatedmcp audit

Scan your MCP configs for risky servers (high/medium/low). Zero auth, instant value.

curatedmcp (no args)

Run as an MCP hub server over stdio for Claude, Cursor, Windsurf, etc.

curatedmcp add <slug>

Add a server from the CuratedMCP catalog to your stack.

curatedmcp remove <slug>

Remove a server from your stack.

curatedmcp list

Show your current stack.

curatedmcp init

Print the config snippet to drop into your AI client.

curatedmcp guard -- <cmd>

Run a server behind the local action firewall.

curatedmcp login

Authenticate the agent to your CuratedMCP account.

curatedmcp sync

Pull your team's registry config and push audit results.


Related MCP server: Agent Factory MCP

1. Audit (the wedge β€” start here)

npx curatedmcp audit

Scans every MCP config file on your machine (Claude Desktop, Cursor, Windsurf, Claude Code, …), classifies each server against the CuratedMCP catalog, and flags:

  • πŸ”΄ HIGH β€” unverified or known-risky servers with credentials

  • 🟑 MEDIUM β€” verified servers running outside catalog defaults

  • 🟒 VERIFIED β€” known-good catalog servers

No signup, no cloud, no data leaves your machine. Logged in? Add --sync to push the result to your dashboard.


2. Run as the MCP Hub

If you use MCP servers across multiple AI clients, you've felt this pain: configure GitHub MCP in Claude Desktop, then re-do it in Cursor, then in Windsurf. New agent ships? Re-paste every config.

The agent fixes that. It's one MCP entry that fans out to every server you've added, in every AI client.

   Claude   Cursor   Windsurf   Copilot   Gemini
       \      \      |      /      /
        β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
        β”‚       curatedmcp         β”‚   ← one config in each agent
        β”‚     (the MCP hub)        β”‚
        β””β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”˜
             β”‚      β”‚      β”‚
          GitHub  Postgres  Stripe   ← `add`'d once, available everywhere

Add it to your AI client

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

Client

Path

Claude Desktop

~/Library/Application Support/Claude/claude_desktop_config.json (mac) / %APPDATA%\Claude\claude_desktop_config.json (win)

Cursor

~/.cursor/mcp.json

Windsurf

~/.codeium/windsurf/mcp_config.json

Claude Code

~/.claude/mcp.json (or .claude/mcp.json per-project)

Add servers to your stack

npx curatedmcp add github          # prompts for GITHUB_TOKEN
npx curatedmcp add postgres --env DATABASE_URL=postgres://...
npx curatedmcp list

Restart your AI client

Tools appear with a <slug>__ prefix:

  • github__create_issue

  • postgres__query

  • filesystem__read_file


3. Guard (local action firewall)

npx curatedmcp guard -- npx -y @modelcontextprotocol/server-github

Wraps an MCP server with a local policy engine that gates every tools/call against ~/.curatedmcp/guard-policy.json. Default policy allows read, prompts on write, blocks destructive.

npx curatedmcp guard --dashboard --port 7878 -- npx -y @some/server
# Then open http://localhost:7878 for the live action log

4. Login + sync (for teams)

Once you have a CuratedMCP account, link the CLI to it:

npx curatedmcp login                  # paste a registry key from your dashboard
npx curatedmcp sync                   # pull team registry config + push audit results
npx curatedmcp sync --team acme-eng   # pick a specific team if you're in more than one

Sync pulls the locked-down server list approved by your team and merges it into your local stack β€” so every developer's machine runs the same vetted set of servers.


Config files

~/.curatedmcp/stack.json β€” your stack, plain JSON, hand-editable, version-controllable:

{
  "version": 1,
  "entries": [
    {
      "slug": "github",
      "name": "GitHub",
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-github"],
      "env": { "GITHUB_TOKEN": "ghp_xxxxxxxxxxxx" },
      "addedAt": "2026-05-01T10:14:00.000Z"
    }
  ]
}

Set "disabled": true on an entry to skip it without removing it.

Other files (created on first use):

  • ~/.curatedmcp/auth.json β€” login token (mode 0600)

  • ~/.curatedmcp/guard-policy.json β€” firewall policy

  • ~/.curatedmcp/launcher.json β€” anonymous client UUID


In-agent discovery

The agent itself exposes discovery tools to your AI client, so you can ask:

"Find me an MCP server for Postgres." "What's the best Stripe MCP?" "Add the Postgres MCP server to my stack."

The agent uses search_servers, get_server_details, and add_to_stack to do all of that without you leaving the chat.


Privacy

  • All config is local at ~/.curatedmcp/. No cloud sync unless you login.

  • Anonymous telemetry only (event names like "search", "add"). Disable with --no-telemetry or CURATOR_TELEMETRY=false.

  • Audit results stay on your machine unless you login and run --sync.


Compatibility

  • Works with Claude Desktop, Claude Code, Cursor, Windsurf, Copilot, Gemini, OpenAI Agents β€” anything that supports MCP over stdio.

  • Node.js β‰₯ 18.


Migrating from the old packages

The agent replaces three earlier packages, which are now deprecated:

Old

New

@curatedmcp/launcher

curatedmcp (no args) / curatedmcp add / curatedmcp list

@curatedmcp/auditor (aka mcp-audit)

curatedmcp audit

@curatedmcp/sentinel (aka sentinel)

curatedmcp guard

A launcher bin alias is kept for back-compat.


MIT licensed.

Available Tools

5 tools
add_to_stackA

Add an MCP server to your Launcher stack so its tools become available through Launcher in every AI client. The server's tools appear as <slug>__<tool> after the AI client is restarted.

Inputs: slug (required), env.

get_server_detailsA

Get full details about a specific MCP server including install instructions

Inputs: slug (required).

install_serverA

Get the manual install configuration snippet for an MCP server (use add_to_stack instead if you want it managed by Launcher)

Inputs: slug (required), client.

list_categoriesA

List all available MCP server categories

Inputs: no named parameters.

search_serversA

Search the CuratedMCP catalog for MCP servers by keyword, category, or use case

Inputs: query (required), limit, category.

Tool Schema Changelog

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

Loading tool schema changelog…

    TDQS

    A4/5.0

    Scored across 5 tools

    Disambiguation4/5

    Tools are mostly distinct: search_servers finds servers, get_server_details retrieves details, list_categories browses categories, while install_server and add_to_stack both relate to making a server available but via different mechanisms. The overlap between install_server and add_to_stack is clarified by their descriptions, so confusion is unlikely.

    Naming Consistency5/5

    All tool names follow a consistent verb_noun pattern with lowercase and underscores: get_server_details, search_servers, install_server, list_categories, add_to_stack. The verbs are clear and the pattern is uniform across the entire set.

    Tool Count5/5

    With only 5 tools, the server is well-scoped for a curated catalog. Each tool covers a distinct core functionβ€”searching, browsing, retrieving details, and two installation pathsβ€”without unnecessary bloat or sparseness.

    Completeness4/5

    The tool surface covers the primary lifecycle of discovering and installing MCP servers: search, filter by category, get details, and install via either manual config or Launcher integration. Minor gaps exist, such as no removal/uninstall tool, but the core workflows are complete.

    Maintenance

    ActivityInactive
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      B
      quality
      F
      maintenance
      A universal MCP server that automatically discovers and registers CLI tools as AI-powered agents with persona configuration, enabling any CLI tool to be used as an MCP tool.
      4
      1
      MIT
    • A
      license
      A
      quality
      B
      maintenance
      An MCP server that bridges AI assistants to other CLI tools, enabling capabilities like large-context reads, vision, file generation, and cross-vendor verification without API keys.
      6
      51 PyPI
      2
      Apache 2.0
    • A
      license
      Not graded
      quality
      C
      maintenance
      Enables orchestrating multiple AI CLI agents (Codex, Grok, Claude, etc.) via Claude Code, providing tools for dispatching tasks, broadcasting, debating, and managing jobs through a local MCP server.
      MIT