MCP Config Generator
Fill in one form, get copy-paste MCP config for every client — Claude Code, Claude Desktop, Cursor, Codex, Gemini CLI and VS Code. Or start from the SkillzBench preset and connect the verified-skills MCP to your agent right now.
The SkillzBench MCP connects your agent to the verified skills you pick — scanned, pinned to a commit and benchmarked. Chat apps (Claude, and any app with remote connectors) just paste the URL and sign in; coding agents use an API key from your keys page.
https://mcp.skillzbench.com/mcpclaude mcp add --transport http \ skillzbench https://mcp.skillzbench.com/mcp \ --header 'Authorization: Bearer sb_live_XXXX'
Chat-app connectors (OAuth, no key): open the connector guide.
MCP setup FAQ
How do I add an MCP server to Claude Code?
For a remote server: claude mcp add --transport http <name> <url> (add --header 'Authorization: Bearer …' if the server needs a key). For a local stdio server: claude mcp add <name> -- <command> <args…>. The generator writes both forms for you.
How do I add an MCP server to Cursor?
Edit .cursor/mcp.json (project-level) or ~/.cursor/mcp.json (global) and add an entry under mcpServers with the server's url and, if required, an Authorization header. The generator produces the exact JSON.
What is the SkillzBench MCP?
One free MCP server that gives any agent access to the verified agent skills you choose — every skill security-scanned, pinned to a commit and benchmarked. Chat apps connect with just a URL and sign-in; coding agents use an API key.
MCP server vs skills — which do I need?
They solve different problems and work together. An MCP server connects an agent to tools and data at runtime; a skill teaches an agent a repeatable procedure. SkillzBench serves its skills through one MCP so you can equip every agent from a single toggle board.
Why serve skills over one MCP?
Installing skills per-agent means drift: different machines, different versions, no visibility. Serving them from one MCP keeps every skill pinned to a verified commit, lets you toggle what each agent sees from one place, and means revoking access is a click instead of a repo sweep. The agent guide covers the full setup, and the methodology explains what "verified" means.