MCP Verdict MCP Server
# MCP Verdict MCP Server
Ask your AI assistant which MCP server to use, and get an answer that comes from an evaluation instead of a guess.
This server gives Claude, Cursor, and any other MCP client access to [MCP Verdict](https://mcpverdict.com): independent, source-reviewed evaluations of MCP servers, client profiles, and setup guides. Reviews cover capabilities, setup, permissions and security, token costs, and maintenance status, with a verdict where one has been published.
Content is fetched live from [mcpverdict.com](https://mcpverdict.com), so answers match the current published reviews. No API key needed.
## Tools
| Tool | What it does |
|---|---|
| `search_servers` | Find reviewed MCP servers by name, category, or capability (`"postgres"`, `"browser automation"`, `"oauth"`). Omit the query to list all. |
| `get_server_review` | Full review of one server, or one section of it: `section: "security"`, `"setup"`, `"cost"`, `"faq"`, `"troubleshooting"`. |
| `get_guide` | Setup guides (Claude Code, Claude Desktop, Cursor), client profiles, and comparisons such as [best MCP servers](https://mcpverdict.com/mcp/compare/best-mcp-servers/) and [MCP vs API](https://mcpverdict.com/mcp/compare/mcp-vs-api/). |
Example prompts:
- "Is the Puppeteer MCP server still safe to use?"
- "Compare the Postgres and Supabase MCP servers for a production database."
- "What token scopes does the GitHub MCP server need?"
- "How do I add an MCP server to Cursor?"
## Install
Requires Node.js 18 or newer.
**Claude Code**
```bash
claude mcp add mcpverdict -- npx -y mcpverdict-mcp
```
**Claude Desktop** — add to `claude_desktop_config.json`:
```json
{
"mcpServers": {
"mcpverdict": {
"command": "npx",
"args": ["-y", "mcpverdict-mcp"]
}
}
}
```
**Cursor** — add to `~/.cursor/mcp.json` (global) or `.cursor/mcp.json` (project):
```json
{
"mcpServers": {
"mcpverdict": {
"command": "npx",
"args": ["-y", "mcpverdict-mcp"]
}
}
}
```
Setup for other clients: [MCP setup guides](https://mcpverdict.com/mcp/setup/).
## How reviews are made
Reviews are researched from official documentation, repositories, release notes, and disclosed external measurements. See the [methodology](https://mcpverdict.com/methodology/) and [editorial independence](https://mcpverdict.com/editorial-independence/) policy. Found an error? [Corrections](https://mcpverdict.com/corrections/).
## Development
```bash
npm install
npm run build
npx @modelcontextprotocol/inspector node dist/index.js
```
Set `MCPVERDICT_BASE_URL` to point at another copy of the site (for example a staging install).
## License
MIT. Review content © MCP Verdict; please link to the source page when quoting it.
TDQS
Scored across 3 tools
The three tools target clearly distinct resources: a server directory search, a guide lookup, and a single-server review fetch. search_servers lists/summarizes while get_server_review drills into one server's details, so boundaries are unambiguous.
All three names follow a consistent verb_noun pattern (search_servers, get_guide, get_server_review) with no mixed conventions. Predictable and readable.
Three tools is on the lean side, but for a read-only review/lookup service the surface is well-scoped and each tool earns its place. Slightly under what a richer browsing experience might want.
Search, guides, and detailed per-server reviews cover the core read-only lifecycle of a review directory, with section-level access for long reviews. Minor gaps like category listing or cross-server comparison are largely absorbed by the search tool.