Asva AI MCP Server

Overview

The Asva AI MCP server exposes your Answer Engine Optimization data to any AI assistant that speaks the Model Context Protocol (MCP). Once connected, the assistant can pull your brand visibility, share of voice, citations, sentiment, competitor rankings, and regional performance on demand, backed by your live Asva AI account.

Authentication runs over OAuth 2.1 with PKCE. Your assistant never sees your password, and you can disconnect at any time from the Connectors page in the dashboard.

Choose your connection method

Pick the setup that matches your AI assistant. All methods connect to the same MCP server and expose the same tools.

Step-by-step setup

Claude Web (claude.ai)

1. Go to claude.ai, click your avatar, then Settings, then Connectors.
2. Click Add custom connector and paste https://mcp.asvaai.com/mcp.
3. Sign in with your Asva AI account when the OAuth popup appears.

Claude Desktop or Claude Code

Run this in your terminal:

claude mcp add --transport http asvaai https://mcp.asvaai.com/mcp

A browser window opens for sign-in. After you authorize, run claude mcp list to confirm.

ChatGPT (Custom Connector)

1. In ChatGPT Settings, open Beta features and enable Custom Connectors. On the Free tier, enable Developer mode under Settings, Connectors, Advanced settings.
2. Go to Settings, then Connectors, then Create.
3. Name it Asva AI, paste https://mcp.asvaai.com/mcp, Authentication OAuth.
4. Sign in with your Asva AI account.

Cursor

1. Press Cmd + Shift + P on macOS or Ctrl + Shift + P on Windows and Linux.
2. Select Open MCP Settings, click Add new MCP server, and paste:

{
  "mcpServers": {
    "asvaai": { "url": "https://mcp.asvaai.com/mcp" }
  }
}

Save. Cursor opens a browser to complete OAuth.

VS Code (Copilot agent mode)

1. Create .vscode/mcp.json in your workspace (VS Code 1.102+ uses the top-level servers key and a type field — do not reuse Cursor's mcpServers shape):

{
  "servers": {
    "asvaai": { "type": "http", "url": "https://mcp.asvaai.com/mcp" }
  }
}

Click Start on the server, authorize in the browser, then pick asvaai from the Tools picker in the Chat view (Agent mode). For a user-level install, run MCP: Add Server from the Command Palette and choose HTTP.

Windsurf

1. Open Windsurf Settings.
2. Go to Cascade, then MCP Servers.
3. Add the configuration below, save, and restart Windsurf.

{
  "mcpServers": {
    "asvaai": { "serverUrl": "https://mcp.asvaai.com/mcp" }
  }
}

How it works

1. 1
   You ask a question in your AI assistant.

   For example, "show me my brand visibility on ChatGPT last week."

2. 2
   The assistant picks the right Asva AI tool.

   It calls your account over an OAuth-authenticated MCP channel.

3. 3
   Asva AI returns your live data.

   The assistant formats the response in chat. Tool calls are audit logged.

What you can do

A few example prompts from customer accounts:

• What is my brand mention rate on ChatGPT for the last 30 days?
• Show share of voice for Gemini versus Perplexity in the US.
• Which of my tracked topics has the worst citation rate this month?
• Generate a visibility audit for stripe.com.
• Drill into the actual AI responses for topic "best analytics tools".
• List my top three improving topics in the last fourteen days.
• Draft a weekly digest email for my team based on the last seven days of Asva AI data.

Tool reference

These are the tools the assistant can call. Every tool respects your account permissions. Read-only tools are marked as such in the MCP metadata; the one writer (generate_visibility_audit) creates a new report row in your own account.

Transport: Streamable HTTP (MCP spec 2025-03-26). All tools authenticate with the same short-lived OAuth 2.1 bearer token and respect the caller's brand permissions. Tool responses carry only analytics metadata about brands the caller has access to — no PII, billing data, or credentials.

Security and privacy

• Every tool call is authenticated with a short-lived OAuth 2.1 bearer token. Refresh tokens expire after 30 days.
• The MCP server never stores credentials or proxies to third-party language models on behalf of the assistant.
• Each call is audit logged with your user ID and timestamp.
• A tool call only returns data for brands the authenticated user has access to. Cross-account reads are rejected at the backend layer.
• You can disconnect at any time from the Connectors page in the Asva AI dashboard. That revokes every token tied to that client.

For the full data-handling policy, see the Privacy Policy. Terms of use are at the Terms of Service page.

Troubleshooting

More documentation