$39/ month
250 credits a month
- 2 brand workspaces
- All public models
- Up to 1,000 personas a sim
- API and SDKs: Python, TS, Rust, Go
- Email support
Pre-flight checkBefore the spend
Ads, headlines, pages, listings, offers, cold emails, ideas. A pinned room of hostile synthetic customers judges every version and tells you what to keep, fix, or kill, and why.
What you can test
Ads, emails, pages, listings, prices. The room reads it the way a stranger would and marks up the exact spot it lost them. Every card below is a real run. Three of them are our own work, and none of them got a keep.
Your audience
Paste your site. Kranth reads it, works out who buys and where they live, then seats people from those cities, each with a budget, a bar for proof and their own way of talking. Below is the room it cast from our own site, and what that room said about our ad.
Debate, live
Put a motion on stage and the room argues it out, turn by turn, in their own voices, while the audience watches and changes its mind. Five formats, from a head-to-head to a pre-mortem at the idea's graveside. Every run below is real.
Interview, live
Interview mode turns the room on you. A panel with something to lose asks what your investor, buyer or boss will ask, scores every answer for evasion, and keeps pushing until you give them a number. This session is real. We played a founder who dodges.
Recon
Recon sends a swarm of research agents across the open web. Each one chases a different question, brings back what it found and pins it to the page it came from. Then an editor tells you what holds, what doesn't, and where the swarm drifted. This run is real.
We ran it on ourselves
Before we cut our free credits, we ran it past 50 personas. They gave it 14 out of 100. We did it anyway. This is what the next nineteen days looked like, with the real numbers.
New signups who could run a first room
How it works
No survey panel and no scraped people. Every seat is assembled from an archetype, a voice and a set of decision traits, then reads your work on its own. This is the same room from the map above, taking our ad.
An ad, a page, a price, an email or an idea. Text or image. Two versions if you want a winner.
Each seat is built from one of 101 archetypes, 9 voice axes and 8 decision traits, then placed in a city your buyers live in.
No groupthink. Each persona reads your work in its own call, without seeing the others. A room can mix models from different labs.
Hostile, neutral and aligned, the objections they share, and which version wins. Then take it to a debate, an interview or recon.
These are synthetic personas. Treat the room as a rehearsal, then go talk to customers.
It won't tell you your conversion rate. It tells you where strangers push back, and why.
Seats are built to disagree. Our own pricing call scored 14.
API
Native SDKs for Python, TypeScript, Rust and Go, a REST API with a live event stream, and a GitHub App that runs a room on a pull request when someone comments /kranth. Below is the exact call behind the room on our map.
Live event stream
/kranth 30Generated by Kranth Reviewer. Comment /kranth to re-run · configure in .github/kranth.yml.
pip install kranthnpm i @kranth/sdkcargo add kranthgo get github.com/Kranth-ai/kranth-goPricing
10 free credits when you sign up, no card. After that, top up from $10 or pick a plan from $39 a month. Credits never expire.
250 credits a month
800 credits a month
4,000 credits a month
Custom credit pools, contractual never-train, dedicated capacity. Not self-serve.
[email protected]10 free credits, no card. Drop the versions, get a keep, fix or kill call before the spend.
Run a free check