Design and code got 10x faster. Customer validation didn’t
Runs inside the tools you already use, nothing new for your team to learn
What’s new in 4.0
Each one turns customer evidence from a project you schedule into something that runs every time you ship.
Each synthetic user starts from a base model trained on 100,000+ hours of real usability sessions, then learns from your customer data. Fast Mode for a quick read, Full Mode for a complete session.
Ask from Claude Design, v0, Replit, Bolt or Lovable. Get what users struggled with — plus a prompt that fixes it.
When a ticket moves to Design Done, a test runs automatically and the evidence lands on the ticket. Jira, Linear, ClickUp and Monday.com.
Build the test, pull the clips that matter, write the summary — in plain language, the way you already work in Claude.
Monitor and score every critical experience across the customer journey, with recommendations for what to fix next.
Your team keeps working where it already works. TheySaid shows up in the canvas, the ticket and the chat — and brings customer evidence with it.
Get feedback on what you just designed — plus a prompt to fix what users struggled with — without leaving the tool.
When a ticket finishes design, a test runs on its own. Feedback and recommended fixes land on the ticket before build starts.
Connect once via MCP and run synthetic user tests from Claude Desktop, Claude Code, the API, or any MCP-compatible client — no separate interface needed.
Synthetic audiences
Synthetic users trained on your customer data really use your product, on screen and out loud. They don’t just click through. They hesitate, get confused, and tell you exactly why.
Side-by-side studies
89%
of issues found by synthetic users
90%
found by human testers
Your own customers, testers we recruit, or third-party panels. Same builder, same Audiences, same report.
The rest of 4.0
Test websites and mobile prototypes on the devices people actually use.
Test iOS and Android apps, with screen and voice recorded as people go.
Invite site or app visitors to give feedback — and control where and how often.
Screening questions that let only your target audience through.
A personal intro video before the session lifts response rate and quality.
AI reviews every video and response, and keeps low-quality ones out of your analysis.
Responses in any language, translated into your account’s language automatically.
Run the same analysis the same way every time — exec summaries, cohort analysis — so results stay consistent from study to study.
User test, interview, survey or poll — one builder, one interface.
Group rating questions on one scale for faster, easier answers.
A redesigned interface built around how builders work now.
Cut the moments that matter and stitch them into a reel stakeholders will actually watch.
Nothing ships on a guess.
