Connect RankMeFast to your AI tools

For the short product walkthrough, see the Connect RankMeFast MCP guide.

RankMeFast has a remote Model Context Protocol (MCP) endpoint. It lets a compatible coding agent, IDE, or CLI read your SEO data and start audits without leaving your current workflow.

Set up in five minutes

1. Open Settings → API keys, select Create key, and copy the key when it appears. RankMeFast shows the full key only once.

2. Use the RankMeFast MCP endpoint:

   https://rankme.fast/api/mcp
   

3. For clients that support environment variables, set the key before starting the client:

   export RANKMEFAST_API_KEY='rmf_REPLACE_WITH_YOUR_KEY'
   

   In PowerShell:

   $env:RANKMEFAST_API_KEY = 'rmf_REPLACE_WITH_YOUR_KEY'
   

4. Choose your client below, paste its configuration, and restart or reload the client.

5. Verify it with: “Use RankMeFast to list my sites.”

Treat the key like a password. Each RankMeFast API key can have its own restrictive scopes. The start_audit tool can consume your monthly audit allowance. Prefer an environment variable, password prompt, or user-level config, and never commit a key to your repository.

MCP access is included with Starter, Pro, and Agency. Agency keys also work with the read-only public REST API.

Control permissions and key scopes

Open Settings → MCP to manage the account defaults shared by MCP and the signed-in AI Assistant:

1. Turn each RankMeFast tool on or off.
2. Leave Allowed sites at all sites, or select the sites tools may use.
3. Turn Allow spending actions off when start_audit must never run.
4. Save the settings. The quick-start card on the same tab shows the endpoint, an API-key shortcut, and a copyable client configuration.

Account defaults are fully permissive until you change them, so existing keys keep working. The Assistant uses these account defaults directly because it runs through your signed-in session.

You can add narrower scopes while creating a key under Settings → API keys, or edit an existing key later. Key scopes can only restrict access. Effective MCP access is always the intersection:

• A tool is available only when both the account and the key allow it.
• When account and key both select sites, only sites present in both lists are available.
• start_audit requires the tool and spending actions to be allowed at both levels.

Leaving a scope unrestricted means “allow the account default”; it never overrides an account restriction. A disallowed tool is omitted from discovery, and a blocked or non-owned site returns the same not-found result.

Language and JSON-RPC contract

For tools/list and requests without a tool override, MCP selects human copy with x-lang, then Accept-Language, then en. Regional header tags such as fr-CA safely resolve to fr. The bearer endpoint never reads browser cookies, account language, or workspace preferences.

Every tool accepts an optional locale argument. An exact value from en, ar, fr, de, es, ru, or zh overrides the headers for that tool’s execution, safe errors, success text, structured-result locale, and HTTP Content-Language. An unsupported argument remains invalid parameters (-32602) in the bearer-request language; it is not regionally normalized. Responses add x-lang, Accept-Language to Vary without removing existing values.

Only tool descriptions, human result summaries, and safe error messages are localized. Tool names and schemas, JSON-RPC jsonrpc, numeric code and id, property names, enum and status values, booleans, counts, IDs, domains, URLs, keywords, timestamps, cursors, evidence, and stored user or provider text stay unchanged. Structured tool results add locale but remove no existing field.

Protocol failures keep their numeric condition: -32700 parse error, -32600 invalid request, -32601 unknown method, -32602 invalid parameters or tool, and -32603 internal error. Their safe message is localized from the numeric code; raw SDK, validation, and vendor diagnostics are never returned. MCP does not emit CSV. The Public API guide documents its byte-stable CSV contract.

Choose your client

Any other client can connect if it supports remote Streamable HTTP and a custom Authorization header. RankMeFast does not expose a local stdio or legacy SSE server.

Claude Code

Add a user-level server. The single quotes keep the environment variable reference intact:

claude mcp add-json --scope user rankmefast \
  '{"type":"http","url":"https://rankme.fast/api/mcp","headers":{"Authorization":"Bearer ${RANKMEFAST_API_KEY}"}}'

Check the connection with:

claude mcp get rankmefast

You can also run /mcp inside Claude Code. The Claude Code surface in the desktop app uses the same configuration. See the official Claude Code MCP guide.

The Claude.ai and Claude Desktop chat connector is different: its remote connector flow documents OAuth, not an arbitrary static bearer header. It cannot connect directly until RankMeFast offers OAuth; use Claude Code instead. See Claude custom connectors.

Cursor IDE and Cursor Agent CLI

Create ~/.cursor/mcp.json for a user-level setup. Replace both placeholders; keep this file private:

{
  "mcpServers": {
    "rankmefast": {
      "url": "https://rankme.fast/api/mcp",
      "headers": {
        "Authorization": "Bearer rmf_REPLACE_WITH_YOUR_KEY"
      }
    }
  }
}

Cursor IDE and Cursor Agent CLI read the same configuration. Verify it with:

cursor-agent mcp list
cursor-agent mcp list-tools rankmefast

For a team configuration, use .cursor/mcp.json but do not put a literal key in that committed file. See the official Cursor MCP guide.

VS Code and GitHub Copilot

Run MCP: Open User Configuration from the Command Palette, then use a password input so the secret is not written into the file:

{
  "inputs": [
    {
      "type": "promptString",
      "id": "rankmefast-key",
      "description": "RankMeFast API key",
      "password": true
    }
  ],
  "servers": {
    "rankmefast": {
      "type": "http",
      "url": "https://rankme.fast/api/mcp",
      "headers": {
        "Authorization": "Bearer ${input:rankmefast-key}"
      }
    }
  }
}

Run MCP: List Servers to start or inspect it. Workspace configuration can live in .vscode/mcp.json; the password-input form is safe to share. See the official VS Code MCP configuration reference.

GitHub Copilot CLI

With RANKMEFAST_API_KEY set in your shell:

copilot mcp add \
  rankmefast \
  --type http \
  --url https://rankme.fast/api/mcp \
  --header "Authorization=Bearer $RANKMEFAST_API_KEY" \
  --tools "*"

The shell expands the key before Copilot saves the user configuration, so protect ~/.copilot/mcp-config.json. See Adding MCP servers to Copilot CLI.

Windsurf / Cascade

Add this to ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "rankmefast": {
      "serverUrl": "https://rankme.fast/api/mcp",
      "headers": {
        "Authorization": "Bearer ${env:RANKMEFAST_API_KEY}"
      }
    }
  }
}

Reload it from Windsurf Settings → Cascade → MCP Servers. See the official Windsurf MCP guide.

Codex CLI, IDE extension, and ChatGPT desktop

Codex CLI, the Codex IDE extension, and the Codex surface in ChatGPT desktop share ~/.codex/config.toml on the same machine:

[mcp_servers.rankmefast]
url = "https://rankme.fast/api/mcp"
bearer_token_env_var = "RANKMEFAST_API_KEY"

Restart the IDE or desktop app after setting the environment variable. In the CLI, check it with:

codex mcp list

You can also run /mcp in a Codex session. See the official Codex MCP guide.

ChatGPT web does not read local Codex configuration, so this setup applies only to Codex surfaces on the configured machine.

Gemini CLI

With RANKMEFAST_API_KEY set in your shell:

gemini mcp add \
  --scope user \
  --transport http \
  --header "Authorization: Bearer $RANKMEFAST_API_KEY" \
  rankmefast https://rankme.fast/api/mcp

Verify it with gemini mcp list or /mcp inside Gemini CLI. The command stores the expanded header in ~/.gemini/settings.json, so keep that file private. In JSON, use httpUrl; Gemini reserves url for legacy SSE. See the official Gemini CLI MCP guide.

OpenCode

Add this to the global ~/.config/opencode/opencode.json file:

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "rankmefast": {
      "type": "remote",
      "url": "https://rankme.fast/api/mcp",
      "enabled": true,
      "oauth": false,
      "headers": {
        "Authorization": "Bearer {env:RANKMEFAST_API_KEY}"
      }
    }
  }
}

oauth: false prevents OAuth discovery because RankMeFast uses a bearer key. Check the setup with:

opencode mcp list
opencode mcp debug rankmefast

See the official OpenCode MCP guide.

JetBrains AI Assistant and Junie

Open Settings → Tools → AI Assistant → Model Context Protocol (MCP), add an HTTP server, and paste:

{
  "mcpServers": {
    "rankmefast": {
      "url": "https://rankme.fast/api/mcp",
      "headers": {
        "Authorization": "Bearer rmf_REPLACE_WITH_YOUR_KEY"
      }
    }
  }
}

Keep this in IDE-level settings rather than a project file. Enable Pass custom MCP servers when you want Junie or another integrated agent to receive the tools. See the official JetBrains MCP guide.

Zed

Open Settings → AI → MCP Servers → Add Remote Server, or add this to your user settings:

{
  "context_servers": {
    "rankmefast": {
      "url": "https://rankme.fast/api/mcp",
      "headers": {
        "Authorization": "Bearer rmf_REPLACE_WITH_YOUR_KEY"
      }
    }
  }
}

Keep the literal key in user settings, not project settings. A green indicator beside the server confirms the connection. See the official Zed MCP guide.

Cline

Open Cline's MCP settings and add a Streamable HTTP server:

{
  "mcpServers": {
    "rankmefast": {
      "type": "streamableHttp",
      "url": "https://rankme.fast/api/mcp",
      "headers": {
        "Authorization": "Bearer rmf_REPLACE_WITH_YOUR_KEY"
      },
      "disabled": false,
      "autoApprove": []
    }
  }
}

Leave autoApprove empty so starting an audit requires your approval. See the official Cline MCP guide.

Roo Code

Use the global MCP settings or .roo/mcp.json:

{
  "mcpServers": {
    "rankmefast": {
      "type": "streamable-http",
      "url": "https://rankme.fast/api/mcp",
      "headers": {
        "Authorization": "Bearer rmf_REPLACE_WITH_YOUR_KEY"
      },
      "disabled": false,
      "alwaysAllow": []
    }
  }
}

Leave alwaysAllow empty so spending actions need approval. See the official Roo Code MCP guide.

Kiro IDE and CLI

Use ~/.kiro/settings/mcp.json globally or .kiro/settings/mcp.json for a project:

{
  "mcpServers": {
    "rankmefast": {
      "url": "https://rankme.fast/api/mcp",
      "headers": {
        "Authorization": "Bearer ${RANKMEFAST_API_KEY}"
      }
    }
  }
}

Kiro expands environment variables in headers. See the official Kiro MCP configuration.

Copilot in other IDEs

GitHub Copilot in Visual Studio, JetBrains IDEs, Xcode, and Eclipse uses a different shape from VS Code:

{
  "servers": {
    "rankmefast": {
      "url": "https://rankme.fast/api/mcp",
      "requestInit": {
        "headers": {
          "Authorization": "Bearer rmf_REPLACE_WITH_YOUR_KEY"
        }
      }
    }
  }
}

Store this as user-level configuration and do not commit the key. Follow GitHub's MCP setup guide for Copilot IDE extensions for the config-file location in your IDE.

Available RankMeFast tools

Reads and actions are always scoped to the account that owns the key. A site, analysis, or run ID from another account is returned as not found.

Your client discovers only the tools allowed by the effective account-and-key permissions. A restricted tool being absent is expected and does not mean the connection failed.

The tool list changes additively: existing tools keep their names and argument shapes, and new tools may appear as features ship. Nothing in this release removes or renames a tool.

Troubleshooting

• The client cannot discover tools: confirm the URL ends in /api/mcp, choose Streamable HTTP rather than SSE or stdio, then restart the client.
• One expected tool is missing: check both the account toggle under Settings → MCP and that key’s scopes under Settings → API keys.
• A site returns not found: confirm that the account and key scopes both allow it. Blocked and non-owned sites intentionally use the same response.
• Spending is not allowed: allow start_audit and spending actions at both levels, or keep it disabled for a read-only client.
• 401 Unauthorized: use the exact Authorization: Bearer rmf_… header. The key may be mistyped, revoked, or attached to an inactive account.
• 402 Upgrade required: MCP needs Starter, Pro, or Agency. See Plans, limits & credits.
• 405 Method not allowed in a browser: this is expected because the endpoint accepts MCP POST requests, not ordinary browser GET requests.
• 429 Too many requests: wait for the rate-limit window shown in the response headers, then retry.
• 503 Unavailable: MCP is disabled or temporarily unavailable on that RankMeFast installation; ask its operator.
• Works locally but not through a proxy: make sure the proxy forwards Authorization, Content-Type, and Accept headers to /api/mcp.

To replace a key, create a new one, update and verify every client, then revoke the old key from Settings → API keys. Revocation is immediate.

Compatibility

The registry in this release is exactly the ten tools listed above. Eight are read-only, set_action_state writes but costs nothing, and start_audit is the only tool that spends a plan allowance. Brand Radar, Review Intelligence, Link Intelligence, Traffic Insights, Keyword Trends, and the SERP Sensor do not have MCP tools. No existing tool name or argument shape changed.

<!-- mcp-tools: list_sites; get_latest_audit_report; list_keywords; get_rank_history; list_content_analyses; get_content_analysis; start_audit; get_audit_status; list_actions; set_action_state -->

Related guides

• AI Assistant, use the same approved tools in RankMeFast chat.
• Public API, the Agency-only read API.
• Content Intelligence, understand the analysis data returned by MCP.
• Account security, keep your account and keys safe.