Skip to content
Intura
AI Workflows & AutomationIntermediate8 min read

Connect Intura MCP to Hermes Agent: AI Visibility and Website Audits from Chat

Add the Data AI Search MCP server to Hermes Agent with one block in ~/.hermes/config.yaml. This guide covers where the API key goes (and the header that does not work), how to test the connection, which tools spend credits, and how to keep the agent from starting audits you did not ask for.

  • #hermes-agent
  • #mcp
  • #intura-mcp
  • #ai-visibility
  • #website-audit
  • #ai-agent-setup
You connect Intura to Hermes Agent by adding one remote server under mcp_servers in ~/.hermes/config.yaml, with your Data AI Search API key in the URL. After a /reload-mcp, the agent can read your AI Visibility reports, Website Audits, Category Overview data and automation results, and act on them in the same chat where it already does your SEO work. As of 20 October 2026 the server has 31 tools in four groups, listed in the table below. That matters if you followed our Hermes Agent SEO series. Guide two gave the agent Search Console and DataForSEO, which tell it how you rank on Google. Neither tells it how Gemini, ChatGPT, Claude or Perplexity describe your brand. The Intura server fills that gap with the audits you already run in the dashboard. Intura publishes this guide and builds the Data AI Search MCP server it describes. The server details come from its own source and the public integration page; the Hermes commands and config keys come from the Hermes Agent MCP and configuration docs, checked on 20 October 2026.

How do you connect Intura MCP to Hermes Agent?

  1. 1

    Copy your API key from the dashboard

    Open Integrations → MCP in the Data AI Search dashboard and copy your API key. The page also shows a full server URL with the key already in it, which is fine for the Claude app but not what you want here: in Hermes you keep the key and the URL apart. Treat the key like a password. Anyone who has it can read the workspace and run audits that spend its credits.

  2. 2

    Save the key in ~/.hermes/.env

    Add the key to ~/.hermes/.env as INTURA_API_KEY. Hermes resolves ${VAR} placeholders in config.yaml from your environment and this file, so the config can name the key without containing it. That keeps the key out of a config file you might paste into a chat, share with a teammate or commit to a dotfiles repo. If the variable is missing, Hermes keeps the placeholder as literal text and the server answers with an authentication error, which is the first thing to check when step 4 fails.

    Terminal
    echo 'INTURA_API_KEY=paste_your_key_here' >> ~/.hermes/.env
  3. 3

    Add the server under mcp_servers in config.yaml

    Open ~/.hermes/config.yaml and add an intura entry with a url and nothing else. The key goes in the URL as ?api_key=, not in a header. This is the step people get wrong: the Hermes docs show remote servers authenticating with headers: Authorization: Bearer, and copying that pattern fails here, because the Intura server reads a Bearer header as an older login token, not as your workspace key. The server speaks Streamable HTTP, which Hermes uses for a url entry by default, so no transport line is needed.

    ~/.hermes/config.yaml
    mcp_servers:
      intura:
        url: "https://api.dataaisearch.com/mcp?api_key=${INTURA_API_KEY}"
  4. 4

    Reload and test the connection

    Type /reload-mcp in a running Hermes chat, or run hermes mcp test intura from the terminal. Hermes prefixes every MCP tool with the server name, so the Intura tools show up as mcp_intura_ followed by the tool name: mcp_intura_audit_list_website_audits, mcp_intura_category_get_questions and so on. If the test fails with an authentication error, the key did not resolve or was copied wrong; go back to step 2. If tools list fine but a call comes back with a credits message, the workspace balance is below 1.

    Terminal
    hermes mcp test intura
    
    # or, inside a running chat
    /reload-mcp
  5. 5

    Decide which tools the agent may call

    Use the tools.exclude list in the server entry to hide the tools that spend real credits until you trust how the agent uses them. Every Intura tool call costs 1 credit, except five that charge what the same action costs in the dashboard: creating an AI Visibility project, adding prompts to it, running prompts, starting a Website Audit and re-running an audit with AI analysis. An agent asked to "check our visibility" can decide a fresh run is the best way to check. Excluding those five keeps the first week read-only in practice; remove them from the list when you want the agent to start runs for you.

  6. 6

    Ask a first question and check which tool ran

    Start with a question that can only be answered from your workspace, such as listing your AI Visibility projects, and watch the tool calls in the Hermes output. If no mcp_intura_ call appears, the agent answered from its own knowledge and the numbers are made up. Guide three in the series covers this habit in more depth; it applies to every MCP server, not just this one.

    Prompt
    List my AI Visibility projects with their latest mention rate. Use the Intura tools only, and tell me which tool you called.

What can Hermes Agent do with the Intura MCP server?

What can Hermes Agent do with the Intura MCP server?
AreaWhat you can askTool names start withCalls that cost more than 1 credit
AI Visibility AuditHow Gemini, ChatGPT, Claude, DeepSeek and Perplexity mention and cite your brand; clusters and next actionsaudit_…_ai_visibility_…Create project, add prompts, run prompts
Website AuditWhat stops AI and Google from reading your pages; problems, pages and action items; re-check one pageaudit_…_website_…Start an audit; re-run with AI analysis
Category OverviewWhat people ask AI in a category you follow: questions, keywords, brands, segmentscategory_None
Automation resultsOutput of the automations you scheduled in the dashboardlist_automation_results, get_automation_resultNone

Copy-ready config and prompts

config.yaml: Intura with credit-spending tools hidden
mcp_servers:
  intura:
    url: "https://api.dataaisearch.com/mcp?api_key=${INTURA_API_KEY}"
    tools:
      exclude:
        - audit_create_ai_visibility_project
        - audit_add_ai_visibility_prompts
        - audit_run_ai_visibility_prompts
        - audit_start_website_audit
        - audit_rerun_website_audit

Exclude takes the tool names as the server lists them, without the mcp_intura_ prefix. Run /reload-mcp after every edit.

Prompt: weekly AI visibility check
Using the Intura tools, get the latest AI Visibility report for <project>. Give me the mention rate and citation rate per AI engine, the three buyer questions where we are not mentioned at all, and the top open next action. Quote the numbers exactly as the tool returned them.
Prompt: fix, then re-check one page
In my latest Website Audit for <domain>, list the open problems on /pricing. I have fixed them. Re-check that page only, wait for the result, and tell me which problems are gone and which remain.

A single-page re-check costs 1 credit. A full re-run with AI analysis costs what it costs in the dashboard.

Prompt: combine with Search Console
For the top 10 queries in Search Console for sc-domain:example.com over the last 28 days, check the Intura Category Overview for matching consumer questions. Which of those questions do AI engines answer without mentioning us?

Needs the mcp-gsc server from guide two in the same config.

Frequently asked questions about Intura MCP and Hermes Agent

How do I connect Intura MCP to Hermes Agent?

Add an intura entry under mcp_servers in ~/.hermes/config.yaml with url set to https://api.dataaisearch.com/mcp?api_key=${INTURA_API_KEY}, put INTURA_API_KEY in ~/.hermes/.env, then type /reload-mcp. The key comes from Integrations → MCP in the Data AI Search dashboard and needs a paid plan.

Can I send the API key in an Authorization header instead?

No. The Intura MCP server reads the workspace API key only from the api_key parameter in the URL. An Authorization: Bearer header is treated as a different, older kind of token, so a workspace key sent that way is rejected.

How many credits does Hermes Agent use on Intura?

Each tool call costs 1 credit, including reads. Five actions charge their dashboard price instead: creating an AI Visibility project, adding prompts, running prompts, starting a Website Audit and re-running an audit with AI analysis. An agent that retries or explores can make more calls than you expect, so watch the balance on the Integrations page during the first week.

Why does hermes mcp test intura fail with an authentication error?

Usually the ${INTURA_API_KEY} placeholder did not resolve, so the literal text reached the server as the key. Check that the variable is in ~/.hermes/.env under exactly that name, with no spaces around the equals sign, then run /reload-mcp. If the key itself is wrong, copy it again from Integrations → MCP; a key from a different workspace connects to that workspace instead.

Can I schedule an Intura report with Hermes cron?

Yes. Once the server is connected, a hermes cron job can call the Intura tools like any other MCP tools, the same way guide two schedules a Search Console report. Keep the credit-spending tools excluded for unattended runs unless you want the job to start new audits, and run the job once by hand with hermes cron run before trusting it.

Key Takeaways

Store the key in ~/.hermes/.env, add one intura entry under mcp_servers with the key in the URL as ?api_key=${INTURA_API_KEY}, and run /reload-mcp. Do not use an Authorization header for this server. Hide the five credit-spending tools with tools.exclude until you trust the agent, check that an mcp_intura_ tool actually ran before you believe a number, and then pair it with Search Console from guide two so one agent sees both how you rank on Google and how AI engines talk about you.