FeaturesMCPPricingBlog
Log inFree Report
Home
/

MCP

MCP

Run the Loop from Claude
Or Grok. Or Cursor.

Measure, research one topic, publish a small cluster, measure again. MCP is how an agent runs that loop — not a second product. Two topical clusters, not thirty articles.

Connect

One hosted URL. Streamable HTTP.

Point Claude, Cursor, Grok, or any MCP client at BrandKarma. Sign in with OAuth when the client supports it. Otherwise send a tenant API key as Bearer.

MCP endpointhttps://app.getbrandkarma.com/api/mcp

Paste this into Claude Code .mcp.json or Cursor .cursor/mcp.json. The client should open a BrandKarma login window. Do not paste a key into a prompt.

OAuth — preferred
{
  "mcpServers": {
    "brandkarma": {
      "type": "http",
      "url": "https://app.getbrandkarma.com/api/mcp"
    }
  }
}
API key as Bearer
{
  "mcpServers": {
    "brandkarma": {
      "type": "http",
      "url": "https://app.getbrandkarma.com/api/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

Grok: grok mcp add --transport http brandkarma https://app.getbrandkarma.com/api/mcp. Create the API key in the BrandKarma app, not in chat.

What you can do today

Jobs, not a catalogue

MCP is how an agent runs the loop. It is not a second product. These tools already ship.

01
See whether AI names you

Ask whether the engines named you, and whether anyone arrived. You can already call get_monitoring_results for mention rank and share of voice, and get_traffic_analytics for sessions and pageviews. Search Console history sits on the same dashboard in the app — sixteen months when you link a property.

You can already
get_monitoring_resultsget_traffic_analytics
02
Research one topical cluster

One topic, researched once: a dossier of cited claims, not a keyword list. You can already call list_brands to pick the brand, start_deep_research to run the job, and get_research_dossier to read the claims.

You can already
list_brandsstart_deep_researchget_research_dossier
03
Approve a small cluster and publish

From that dossier: a pillar and three to five supporting pieces. Approve the work, write from the evidence, and publish. Pull the published pieces through the content API and render them in your stack. This site’s blog runs on it.

You can already
publish_workstream
Honest cost

Research spends a cluster. Talk does not.

Reading analytics, monitored prompts and brand config does not spend a cluster. Research spends a cluster from the plan — two a month on Brand. Writing the pillar and supporting pieces from that dossier is the same cluster, not a new mill. Extra cluster is $19 this period.

What This Is Not

Not a rank tracker
The point is the join: what the engines say, next to whether anyone arrived. MCP is how an agent runs that loop.
Not BYOK DataForSEO
Keyword intelligence is included. You do not bring your own rank-tracker keys to make the product work.
Not a monitor-only SKU
Measurement without a way to change the number is a vanity metric. The cluster is the intervention.
Not a content mill
We will not help you flood. Thin, duplicate and “Complete Guide to X” ideas are rejected.

Questions People Actually Ask

See What AI Says About You

Request a free AI Visibility Report. We run your top categories across ChatGPT, Perplexity and Gemini and send it within 48 hours. Then start Brand and point Claude at the same loop.

Request Your Free Report