Documentation

MCP: run Ansengine from an AI assistant

46 tools in Claude, ChatGPT, Cursor and Claude Code, with the same explicit confirm a human gives.

The MCP server is the whole product without the dashboard. It is the same API your console uses, so plan limits, roles and budgets apply identically.

Connect it

1. 1Mint an API key under Settings, then API. API and MCP access are included from the $79 tier up.
2. 2Add the server to your client with the key as a Bearer token.
3. 3Call list_brands first: every other org tool takes a brandId from that list.

{
  "mcpServers": {
    "ansengine": {
      "url": "https://mcp.ansengine.co/mcp",
      "headers": { "Authorization": "Bearer ak_your_key_here" }
    }
  }
}

What agents can do

• Read everything the dashboard shows: visibility, trends, competitors, sources, the funnel, prompts, locations, receipts, the QBR, budget status.
• Configure: add or archive prompts, set competitors, seed the outreach worklist.
• Act: run scans and audits, capture a local grid, draft content and pitches, publish to your CMS, decide approvals, send outreach, convert a prospect to a client, ask Pulse.

The confirmation step

Any tool that spends budget or cannot be undone pauses on the first call. It returns a plain summary of what is about to happen plus a short-lived confirmation token bound to your workspace, that tool, and those exact arguments. Calling again with the token executes it. This is the agent-shaped version of the console's two-click confirm: an agent needs the same explicit yes a human does.

> send_outreach_pitch (brandId, targetId)
CONFIRMATION REQUIRED (this is irreversible).
SEND the drafted pitch from the org's connected Gmail. This emails a real
person and cannot be unsent.
Call again with the SAME arguments plus confirmToken. Expires in 10 minutes.

> send_outreach_pitch (brandId, targetId, confirmToken)
Sent from your Gmail. Status: sent.