Claude Code setup guideData AI Search × Claude Code

Add Data AI Search to Claude Code

One command adds the Data AI Search MCP server to Claude Code over HTTP. Claude Code can then read your brand's Insights and audits while it writes pages, fixes what the Website Audit found, or drafts content around the questions buyers really ask.

Get your API keyHow the MCP server works

Setup01

Connect Claude Code to Data AI Search

  1. 1

    Copy your server URL

    Sign in to the Data AI Search dashboard, open Integrations → MCP and copy the server URL. Your API key is already in it.

  2. 2

    Add the server

    Run the command below, with the URL in quotes. --transport http tells Claude Code it's a remote server; --scope user makes Data AI Search available in every project on this machine. Use --scope local to keep it to the current project.

    Terminal
    claude mcp add --transport http --scope user intura \
  "https://api.dataaisearch.com/mcp?api_key=YOUR_API_KEY"
  3. 3

    Check that it's connected

    Run claude mcp list, or type /mcp inside Claude Code. Data AI Search shows up as intura with its three tools.

    Terminal
    claude mcp list

# or, inside Claude Code
/mcp
  4. 4

    Optional: allow the read-only tool

    Claude Code asks before every MCP call. To stop it asking for brand_insights, which only reads, add the tool to permissions.allow in your settings. Leave the audit tools out, so runs that spend credits still need your OK.

    ~/.claude/settings.json
    {
  "permissions": {
    "allow": ["mcp__intura__brand_insights"]
  }
}

Replace YOUR_API_KEY with your own key, or copy the full URL from Integrations → MCP in the Data AI Search dashboard, where it is already filled in. The URL is the key: anyone who has it can read your workspace and spend its credits.

What it can reach02

What Claude Code can do with your data

The server has three tools, one per pillar. Each takes a brand name; ask about a brand the workspace doesn't have and the tool replies with the names it does have.

  • brand_insights

    Read-only. The real questions your brand's buyers ask AI, how AI answers them, who it names instead of you, keywords and segments.

    Tool reference

  • ai_visibility_audit

    Read the Visibility Score, prompts, clusters and next actions, or set the audit up, add questions and run them.

    Tool reference

  • website_audit

    Read the site audit, root causes, weakest pages and action list, or start, refresh, re-run and re-check a page.

    Tool reference
Try these first
  • “Use the intura tools: which pages on [brand]'s site have the biggest Website Audit problems?”
  • “Pull the top discovery questions for [brand] from Data AI Search and draft an FAQ section that answers them.”
  • “Read the open Website Audit actions for [brand], fix the ones in this repo, then re-check the pricing page.”
  • “Compare [brand]'s mention rate on Gemini and DeepSeek in the latest AI Visibility report.”
Good to know03

Before you rely on it

  • To share the setup through a repository, commit a .mcp.json whose URL ends in ?api_key=${INTURA_API_KEY}. Claude Code fills in environment variables, so each person sets their own key and none ends up in git.
  • The tools appear as mcp__intura__brand_insights, mcp__intura__ai_visibility_audit and mcp__intura__website_audit. Use those names in permission rules.
  • Claude Code suits Website Audit work well: it can read the problem list, change the code, then ask the server to re-check the page it fixed.
  • To remove the server, run claude mcp remove intura.
If it doesn't connect
claude mcp list shows intura as failed
Add it again with the URL in quotes, exactly as copied. Unquoted, the ? in the URL breaks in most shells.
401 or "Authentication required"
The key has to be in the URL as ?api_key=. This server doesn't read an API key from a header.
"MCP integrations aren't available on your plan"
The workspace is on the Free plan. Upgrade to Starter or higher.
FAQ04

Claude Code and Data AI Search: questions

Which transport does it use?

Streamable HTTP. Add the server with --transport http; there's no local process to run and no SSE setup.

Can my team share one configuration?

Yes, without sharing a key. Commit .mcp.json with ${INTURA_API_KEY} in the URL and have each person export INTURA_API_KEY with their own key.

What does it cost?

Each call uses your workspace's credits: 1 credit per call, except setting up or running AI Visibility prompts and starting a Website Audit, which cost what they cost in the dashboard.

Setup guides for other assistants