Browse docs
Getting started
Setup
Workspaces
Team
Prompts
Brands
Models & coverage
Billing
Results
Metrics
Chats
Sources
Reports
API & MCP
What's new

MCP recipes

Four things to ask your AI assistant once BrandAxis is connected — your best prompts, citation gaps, negative answers and period-over-period change — and the tools it uses for each.

Updated

Once BrandAxis is connected to your assistant, you can ask in plain English. Each recipe below gives a question to ask and the tools the assistant uses to answer it, which helps if you want to steer it or check its work.

Find your strongest prompts

Ask: “Which of our prompts give us the highest visibility, and on which models?”

  1. listWorkspaces finds the workspace.
  2. listBrands finds your brand’s id.
  3. getBrandReport with dimensions: ["prompt"], your brand in brands, and order: "visibility.desc".

Add "model" to the dimensions to split each prompt by model.

Find sites that cite competitors but not you

Ask: “Which sites do AI answers cite most often that never mention us? Show me the pages.”

  1. Find the workspace and your brand’s id.
  2. getSourceReport with kind: "domains", brand: <your brand id>, mentioned: "no" and order: "chats.desc".
  3. Drill into a site’s pages with kind: "urls" and domain_id: <site id>.
  4. getPageContent reads a promising page, so the assistant can say what it covers and where you could fit in.

These are the sites to pitch, get listed on or publish on.

Look into negative answers

Ask: “Where is sentiment towards us weakest, and what are those answers actually saying?”

  1. getBrandReport split by prompt and model finds the weak spots.
  2. listChats filtered to those prompts and models, with your brand and include: "response", pulls the answers.
  3. getChat reads one in full: the brands it names, their position and sentiment, and the pages it cited.

Compare with the previous period

Ask: “How did our visibility change this month compared with last month, by model?”

Any report tool accepts compare: "previous". Each row then carries a previous block for the window of the same length just before, plus previous_from and previous_to. Ask for a breakdown by model, prompt, topic or country to see where the change came from.