Sign inGet started
Your agent (MCP)

Use cases

View as MarkdownOpen in ChatGPTOpen in Claude
Connect your agent
Add to CursorAdd to VS Code

Questions to ask your agent, by task, with what comes back, and multi-step recipes that build on each answer.

Updated

The quickest way to learn what your agent can do is to ask it something. Below are questions that work, by task, each with what comes back. Copy one, change the site's name, and ask in your own words; your agent picks the tools.

Single questions

How is the site doing?

  • "How did shop.example do this week against last week?": visitors, page views and conversions, with the change and the three biggest movers (get_dashboard, find_changes).
  • "Who's on the site right now?": visitors in the last five minutes and their pages (get_live).
  • "Anything I should know this week?": signals from our rules and findings from your skills (list_insights).

Where do visits come from?

  • "Visits by channel this month, and which one grew most.": a table by channel with the change (query).
  • "How much traffic comes from ChatGPT, Perplexity and Claude, and to which pages?": AI assistants are their own channel (query).
  • "Which campaigns bring visitors who sign up?": conversions by campaign (query).

Why did something change?

  • "Why did visits drop on Tuesday?": the channel, page or campaign behind it (find_changes).
  • "What happened around the drop on the 12th?": launches, campaigns, outages and tag changes on the timeline (get_timeline).

What should I fix?

  • "Which pages do people leave right away?": pages by bounce rate, with visits (query).
  • "Where do people give up on the contact form?": the field where most people stop (query).
  • "Which pages are slow on mobile?": Web Vitals by page (query).

Search and ads

  • "Which Google searches almost reach page one?": queries in positions 8 to 15 with their impressions (seo_queries).
  • "Which pages rank in Google but get few clicks?": pages with a low click-through rate (seo_pages).
  • "What does each campaign return per euro spent?": return on ad spend by campaign (query).

People and revenue

  • "Who visited pricing three times but never started a trial?": the people, after consent (find_people).
  • "Save them as a cohort called Pricing lurkers.": a cohort to come back to (save_cohort).
  • "How many leads did the Black Friday campaign bring in Odoo?": leads and won deals by campaign (query).

Your team's context

  • "Remember that our busy season is December.": a note in the team's memory, after you agree (remember).
  • "Note that we raised prices on November 3.": an event on the timeline (add_note).
  • "Count a sent contact form and a visit to /thank-you as conversions.": the project's conversions (set_conversions).

Acting through your tools

  • "Open a Linear issue for the broken checkout form.": through a connector your team added (call_connector).

Recipes

Each step builds on the answer before it, so ask them in order, in one conversation.

A drop, from first look to a fix

  1. "How did shop.example do this week?"
  2. "Why did sign-ups drop on Tuesday?"
  3. "Show me the pages behind it on mobile."
  4. "Open a Linear issue with these numbers, and note the drop on the timeline."

A monthly ads review

  1. "What did each ad campaign spend last month, and what did it return?"
  2. "Which keywords spent the most with no conversion?"
  3. "Link the Google Ads campaign Spring Sale to utm_campaign spring_sale."
  4. "Add the Wasted ad spend skill so this runs every day."

Search wins this week

  1. "Which Google searches almost reach page one?"
  2. "For the top three, read the page and suggest a better title under 60 characters."
  3. "Change the titles in our repository and show me the diff before committing."

Keep exploring

Related
  • Your agent: Connect your own agent to nextanalytica's MCP server and ask about your site in plain words, instead of reading a dashboard.
  • What is an analytics MCP server?: An analytics MCP server lets your agent read your site's visits and answer questions about them. What it is, what it can do and how to connect one.
  • MCP tools reference: Every tool your agent can call on the nextanalytica MCP server, with its arguments and example prompts.
  • FAQ and advanced setup: What people ask before they connect an agent: cost, clients, keys, projects, where data goes and safety, answered straight.
  • Agent keys and OAuth: Two ways your agent signs in as you: the browser sign-in, or a personal agent key for a server or script with no browser, and which to use.
  • Team access: Who on your team can connect an agent to nextanalytica, what the owner, admin and member roles allow, and how a person's or a key's access ends.
Still have questions?
Was this page useful?