Ask your AI agent about your AI search visibility

• Claude Code
• Cursor
• VS Code
• Any MCP client that speaks Streamable HTTP

Claude Desktop only loads local servers, so it connects through a local stdio bridge.

{
  "mcpServers": {
    "ziptie": {
      "url": "https://api.ziptie.ai/mcp",
      "headers": {
        "Authorization": "Bearer zt_live_..."
      }
    }
  }
}

The wrapper key and the type field differ between clients; the client guides show each exact shape.

• Which competitors gained citations this week?
• Which prompts lost visibility over the last 30 days?
• Find answers that state wrong facts about our brand.
• Add these 10 prompts and tag them pricing.
• Which of our pages do AI engines cite, and how are their Search Console clicks trending?

• Projects

  List your projects, read one in full and check the limits that apply to it.

• Prompts and checks

  List, create and update prompts, and read each prompt's performance and every check and answer behind it.

• Competitors

  The competitor roster, mentions, citations, sources and share trends, and competitor tags.

• Citations and sources

  The domains and pages AI engines cite for your prompts, and the source mix.

• Dashboard trends

  The overview and the trend lines for each engine over a window you choose.

• Search Console

  Daily clicks and impressions, plus the pages and queries Google reports.

• Website pages

  The pages of your site that engines cite, with their Search Console clicks beside them.

• Brand facts

  Register the facts about your brand and read the facts, aliases and competitor facts on file.

• Answer accuracy

  The answers that contradict a registered fact, with the claims in the engine's own words, and a way to dismiss a wrong flag.

• Tags

  Create, update and apply tags to prompts and URLs.

• One project or the whole workspace

  A project token reaches one project. A workspace token reaches every project in the workspace, including projects you add later. A token cannot be limited to a chosen set of projects.

• Read, write or admin

  Read lets the assistant look. Write also lets it create and update. Admin also lets it delete. A token carries exactly one scope, fixed when you create it.

• Start with a read-only token

  Delete tools are real deletes, with no confirmation step and no undo. Use a read token to explore, and mint a write or admin token only for the work that needs it.

Rate limits are 60 requests a minute for a project token and 120 for a workspace token. Revoke a token at any time and it stops working on the next request.

• JSON only

  Tools return JSON. There are no file downloads such as CSV, Excel or PDF.

• Poll-based

  Your assistant reads data when it asks. There are no webhooks and no streaming updates.

• No sandbox

  A token acts on your real data, so a write changes your live project.