MCP is how Cursor, Claude Desktop, and similar hosts expose tools to an assistant. RankWatch’s MCP server turns your existing API key into something those assistants can call—workspaces, projects, credits, and tools—without you building a custom plugin for every IDE.
RankWatch already has a full HTTP API and playground. MCP is not a replacement—it is the bridge for people who already work inside an AI host and need RankWatch data in that loop, with the same enterprise controls you expect from any integration.
MCP lets the assistant call RankWatch tools the same way it calls a filesystem or database. Your team describes the outcome; the host handles the tool round-trips.
Nothing special-case. You still mint an organization API key, pick scopes, and optionally restrict workspaces. Revoke the key and the assistant loses access immediately.
Cursor and Claude are already on many desks. MCP meets people where they work instead of forcing another custom dashboard or brittle browser extension.
These are patterns we see with agencies and in-house SEO platforms—not theoretical demos.
Your SEO leads already jump between Cursor and RankWatch. With MCP, they can ask “which workspaces am I allowed to see?” or “start a brand prompt job for this domain” without leaving the editor—or pasting screenshots into Slack.
If you are standardizing on Claude Desktop or another MCP host for internal assistants, RankWatch becomes one more governed tool in that toolkit: same org keys, same scopes, same audit story as any other API integration.
Credit remaining, project lists, tool entitlements—routine checks that used to mean opening the app or writing a one-off script. An MCP-aware assistant can fetch them in the conversation where the decision is happening.
Before you commit an SDK or a custom middleware, MCP is a fast way to explore what the RankWatch API can do with a real key. Treat it as a discovery surface, then harden the path you keep.
Four steps. If you already have a RankWatch API key, you can be live in a few minutes.
Organization owners create keys in Organization settings at go.rankwatch.com when the plan includes API access.
Prefer least privilege: workspaces and projects for read-only assistants; add tools:write only when you want the assistant to start jobs.
Point Cursor or Claude at @rankwatch/mcp and pass RANKWATCH_API_KEY. The server uses https://go-api.rankwatch.com — no host to configure.
Start with account or workspace list, then try a tool job if your key allows it. Keep the secret in the MCP host’s env config, not in chat history.
export RANKWATCH_API_KEY=rw_live_… npx @rankwatch/mcp
{
"mcpServers": {
"rankwatch": {
"command": "npx",
"args": ["-y", "@rankwatch/mcp"],
"env": {
"RANKWATCH_API_KEY": "rw_live_…"
}
}
}
}Each tool maps to the RankWatch API and respects the scopes on your key. Grant only what the assistant should be allowed to do.
| Tool | Scope | Purpose |
|---|---|---|
| rankwatch_get_account | Any valid key | Org context and key metadata |
| rankwatch_list_workspaces | workspaces:read | What the key can see |
| rankwatch_get_credits / limits | usage:read | Capacity before you kick off work |
| rankwatch_list_projects / get_project | projects:read | Project inventory for a workspace |
| rankwatch_tool_entitlements | tools:read | Which tools are available |
| rankwatch_create_tool_job | tools:write | Start a RankWatch tool run |
| rankwatch_get_tool_job / result | tools:read | Poll status and read output |
Store the key in the MCP host environment, not in prompts. Use workspace allowlists when an assistant should only see certain clients. Rotate or revoke from Organization settings the moment a laptop or automation is retired. For deeper API work, use the OpenAPI reference and SDKs; MCP is the conversational front door, not the only door.
Need a commercial or partner integration beyond MCP? Contact partnerships at support@rankwatch.com.