Analyze AI - AI Search Analytics Platform

Agents

What Agents are

A visual builder for automated workflows over your visibility data — chain data recipes, AI steps, logic, and integrations into agents that run on their own.

Automation with your data built in

Agents are workflows you assemble on a canvas: a Start node takes inputs, steps transform them, and an End node returns outputs. What makes them different from generic automation tools is what's available as building blocks — your entire Analyze AI dataset, exposed as data recipes (competitor gaps, share of voice, citation landscapes, brand vault content and dozens more), sitting alongside AI steps, logic, and integrations.

The agent builder: a canvas with Start and End nodes, a searchable steps library including Notion and HubSpot actions, and an input configuration panel with types like short text, JSON, knowledge base, data recipe, and query set
The builder: canvas in the middle, step library on the left, and typed inputs — including data recipes — on the right.

The building blocks

  • Data recipes — pre-built queries over your workspace: brand vs competitor, competitor message shift, ranked keywords, citation breakdowns, and the rest of the catalog. Your data arrives in the workflow already shaped.
  • AI steps — prompt an LLM with your choice of model, with workflow variables interpolated into the prompt, plus purpose-built steps like research-and-plan.
  • Logic — conditionals, branches, and loops for workflows that decide things.
  • Integrations — steps for tools like Notion and HubSpot (query databases, create contacts and deals, append content), plus API calls, code steps, PDF export, and email.
  • Utilities — waits, date formatting, UTM building, text transforms.

Every published agent shows its run history with duration and cost per run, so automation stays accountable.

A published Content Writer Agent: a Start node carrying two data recipes — Brand vs Competitor and Competitor Message Shift — flowing into a Prompt LLM step and a research-and-plan step, with the last run costing $0.0193
A published agent: two data recipes feed an LLM step and a research planner — last run: 119 seconds, $0.02.

What people build with them

  • Recurring competitive reports — recipes → LLM synthesis → email or Notion, on a schedule.
  • Content pipelines — competitive data → research and plan → hand-off to the Content Writer.
  • Alerts with judgment — data recipe → conditional → notify only when something's actually worth a human's time.
  • CRM enrichment — visibility intelligence pushed into HubSpot where sales can see it.

Ready to build one? Start with building an agent, or steal from the examples.