The best prompt is the one you didn't have to leave your editor to write. MCP makes the optimizer addressable from the same chat surface your AI assistant already lives in — no copy-paste shuffle, no second window.
What you get when you wire it in.
One npx command
No install, no build step. Drop the JSON in your IDE config and the optimizer is wired into the tool palette on the next restart.
API-key auth, scoped per-user
Generate a key in /settings#developers. It signs every MCP call — same rate limits and audit trail as the web app, no shared service account.
Same optimizer as the web
MCP routes to the same prompt-optimization service as the web Studio. Whatever the team ships there, your IDE picks up automatically.
Works with any MCP client
Claude Code, Cursor, VS Code today. Any future MCP-aware editor works the same — that's the whole point of the protocol.
Three editors. Same JSON.
Pick your editor, paste the config, restart. The PromptPilot tool appears in the editor's MCP tool palette.
~/.claude/mcp.json{
"mcpServers": {
"promptpilot": {
"command": "npx",
"args": ["-y", "@abhi1381/promptpilot-mcp"],
"env": { "PROMPTPILOT_API_KEY": "pp_..." }
}
}
}Cursor → Settings → MCP{
"mcpServers": {
"promptpilot": {
"command": "npx",
"args": ["-y", "@abhi1381/promptpilot-mcp"],
"env": { "PROMPTPILOT_API_KEY": "pp_..." }
}
}
}settings.json — mcp.servers"mcp.servers": {
"promptpilot": {
"command": "npx",
"args": ["-y", "@abhi1381/promptpilot-mcp"],
"env": { "PROMPTPILOT_API_KEY": "pp_..." }
}
}Wire it into your editor in 60 seconds.
Grab your API key from Settings → Developers, paste the config, restart your editor. That's the install.