# Zaplens

> Synthetic UX intelligence. LLM persona agents browse your website at
> scale, follow goals you describe, and produce a plain-English report on
> where real users would get confused, attracted, or drop off. Used
> instead of (or before) week-long usability studies, multivariate A/B
> tests, and intern click-throughs.

Zaplens is operated by Quantum One AI, a company registered in India
with its principal office in Kolkata. There's a free Hobby tier with a
small monthly credit grant, plus paid Plus ($29/mo) and Pro ($99/mo)
plans. Pro unlocks the MCP server so you can run simulations from
Claude / ChatGPT / Cursor.

## What it does

- **Persona agents** — bring your own personas as markdown, or use the
  default matrix (8 dimensions of realism: confidence, device,
  accessibility, intent, etc.).
- **Heatmaps + drop-off paths** — overlap of click / confusion /
  attraction / repulsion channels on every page, with the agent's
  reasoning attached to each hot spot.
- **Battle mode** — point two URLs at the same goal, watch the agents
  pick a winner. Useful for landing-page variants and pricing-page
  experiments.
- **Exec report** — top drop-offs, persona-specific failures,
  prioritized fixes — written as paragraphs, not as raw charts.

## Core documentation

- [API reference](https://zaplens.site/docs/api): REST
  endpoints for simulations, battles, agents, reports, share links.
- [MCP setup](https://zaplens.site/docs/mcp): how to install
  and authenticate the official MCP server (pipx install zaplens-mcp)
  with Claude Desktop, Cursor, and Claude Code.

## Legal & support

- [Terms of Service](https://zaplens.site/terms)
- [Privacy Policy](https://zaplens.site/privacy)
- [Refund & Cancellation Policy](https://zaplens.site/refund)
- [Contact](https://zaplens.site/contact)

## Not affiliated with

Zaplens is independent from Anthropic, OpenAI, Google, Microsoft, and
any LLM provider. We use LLMs internally to power the agents, but
Zaplens itself is the product.