Docs · MCP server

Let your AI agent do your BrandReco actions

BrandReco's MCP server at https://brandreco.com/mcp lets Claude, ChatGPT, Claude Code or Cursor read your actions, make the change, mark it done and verify it went live. Every action carries a brief written for an agent: the goal, the exact page or file, and how BrandReco will check it. Included on every plan.

Connect your client

Claude (web and desktop)

Open Settings → Connectors and add a custom connector.
Name it BrandReco and paste the URL https://brandreco.com/mcp.
Connect: Claude sends you to a BrandReco consent screen, where you pick the workspace and whether the agent may complete actions.

ChatGPT

In Settings, open Connectors and create a new connector (where your plan requires it, turn on developer mode first).
Use https://brandreco.com/mcp as the MCP server URL, with OAuth.
Approve the BrandReco consent screen. In a chat, enable the connector and ask away.

Claude Code

claude mcp add --transport http brandreco https://brandreco.com/mcp
# then, inside Claude Code: /mcp  →  brandreco  →  Authenticate

Or with a token from Settings → Connections

claude mcp add --transport http brandreco https://brandreco.com/mcp \
  --header "Authorization: Bearer brmcp_your_token_here"

Cursor (and other JSON-configured clients)

Create a token in BrandReco under Settings → Connections, then add this to ~/.cursor/mcp.json (or your client's MCP config):

mcp.json

{
  "mcpServers": {
    "brandreco": {
      "url": "https://brandreco.com/mcp",
      "headers": { "Authorization": "Bearer brmcp_your_token_here" }
    }
  }
}

OAuth clients need nothing but the URL. Tokens are created under Settings → Connections, shown once, and revocable any time; BrandReco stores only a hash. A read token can look; a write token can also start, complete and validate actions — nothing else.

A worked example

In Claude Code, inside your site's repository, with the server connected using a write token:

YouList my BrandReco actions for today and do the first one.
Agent → list_actions{ "today_only": true } — three actions, most important first.
Agent → get_actionReads the first one's steps and its agent_prompt: the goal, the exact file, and how it will be checked.
Agent → start_actionMarks it in progress.
AgentEdits the file in your repository (for example public/llms.txt) and you deploy.
Agent → complete_action{ "note": "Published llms.txt" } — BrandReco starts re-checking your site.
Agent → validate_actionRe-fetches the page and re-runs the check: { "applied": true }. The action shows as verified in BrandReco.

Tools

Read

`get_brand_overview`	The brand being tracked, its competitors, how many actions are open, and page-type coverage versus competitors.
`list_actions`	Open actions, most important first. today_only: true returns today's set; category and limit narrow it.
`get_action`	Everything about one action: why it matters, the exact steps, paste-ready content, and agent_prompt — a complete brief the agent can execute.
`get_visibility`	How the AI agents actually answered your tracked questions, with sentiment and cited sources.
`get_cited_sources`	The websites those answers were built from — yours, competitors', and third parties'.
`get_competitor_gaps`	Page types your competitors have and you don't, and the ones where you're ahead.

Act — write tokens only

`start_action`	Mark an action as in progress before doing the work.
`complete_action`	Mark an action done once the change is made, with an optional note. BrandReco starts re-checking your site straight away. Idempotent.
`validate_action`	Re-fetch your site and re-run the original check now; returns applied: true when the fix is live.

Free covers the action tools; Pro adds the measurement tools (visibility, cited sources, competitor gaps). Everything is scoped to your account, and to one workspace if you pin the token there.

New here? Run the free AI visibility check or read why it matters.