See the experience unfold.
Critical Friends explore while clicks, reactions, friction, and asks arrive. Your agent follows the sessions and brings the evidence back into your workflow, so you hear about it from them instead of from churn.
Coverage from your agent
Run the whole loop
from the agent you already use.
Crities are the synthetic users who try your product. Your AI agent, whether that is Claude Code, Codex, Gemini, or another, is how you direct them: it connects your product, adds the jobs worth covering, launches sessions, reads the evidence, and implements product improvements.
You set the goal and make the calls. Your agent takes on the setup and the work.
Start in your agent
Your product.
Your setup prompt.
We’ll discuss your product and help you get connected during your pilot. Your tailored setup prompt guides your AI agent through the connection and first session.
Start a pilotStart a pilotSet up your productCopy your prompt
Codex · Claude Code · Gemini CLI · Other MCP agents
Already have an account? Sign in ↗From feedback to autonomous action
Give your agent
an experience to improve.
Keep the improvement cycle moving.
Delegate a scope of work. Your agent can investigate findings, make the change, and send the Crities back within that scope, while you stay on your ideal customer.
Choose what better means.
Set priorities, steer the experience, and decide which tradeoffs fit your product. Your agent implements the changes. Your taste shapes the direction.
Covered as you ship
Wake up to
what changed.
When a change breaks a job you cover, you hear about it, grouped by what repeats and who it affects, so you spend your judgment on the few findings that matter, then hand the change to your AI agent.
Since yesterday’s changeInvite a contractor · 48 illustrative sessions across 5 personas
MK
Maya First-time owner
“Do I need to buy a seat first?”
AS
Arun First-time owner
“I’ll check billing before inviting.”
SR
Sam Returning admin
“I’ll add them as a project guest.”
+45 more sessions from Crities like them
Related friction, grouped29 first-time owners take the same billing detour.
Returning admins mostly go straight to the invite, which sharpens the contrast.
Your AI agent acts. Critical Friends take another look.
Build ↔ RevisitAutonomous cycles. Your direction.
Repeated pattern · Invitation flow
New owners mistake inviting for a billing decision.
31 of 48 sessions18 CritiesStrong coverage
What comparing sessions reveals
- First-time owners
- 29 of 32 looked at billing before inviting
- Returning admins
- 2 of 16 took that detour
Consistent across five runs. The retest will show whether the fix holds for both groups.
Suggested action · Your decision
Make guest access clear before seat pricing.
Help new owners find the invitation path while preserving the direct route returning admins already use.
Then test both groups.Do new owners avoid the detour? Can returning admins still move quickly?
Each signal stays linked to its session.Inspect an example finding ↗
Step into the moments that matter
Be pulled into
interesting sessions.
Let your agent keep the work moving. When it surfaces a surprising reaction, stubborn friction, or a product tradeoff, step into the session with the context to unpack it.
An interesting moment
“Do I need to manage the team or buy seats first?”
Maya’s question points to a decision about the invitation experience. Follow the moment to understand what led her there.
Unpack the example sessionThe context comes with it.
- Who’s trying
- The Criti’s perspective, goal, and starting account state.
- What happened
- The screen, clicks, and quoted impressions around the moment.
- Why it matters
- The friction or ask, its supporting evidence, and the finding it informs.
Your agent completes delegated work while you go deeper where your judgment adds value.
Your agent handles the setup
Paste once.
Let the conversation guide you.
Start in your project. Your agent helps connect your Critical Friends workspace, choose the first experience, and prepare the right starting state.
Start a pilot ↗Connect through MCP.
Your agent uses scoped access to plan sessions, follow progress, and retrieve findings and evidence.
Prepare the world they’ll explore.
For signed-in journeys, your agent can implement application hooks that prepare realistic test accounts and data.
Keep the loop in your workflow.
Give the agent your priorities and implementation boundaries. It can implement changes based on the findings and check the experience again.