Ask the users you can't recruit.
Give your Claude, Cursor, or opencode a panel of research personas. Describe your customer in one sentence; get reactions with support counts, severity, and verbatim quotes — before you ship.
$ you · "solo founders building B2B SaaS in Europe, technical, 1-10 employees" [->] create_cohort("Solo SaaS Founders EU Clean")[ok] 5 personas matched · ch_k5705…8derm6 [->] simulate_survey("react to this pricing page")[ok] bundle bdl_e5d68fc3 compiled · your host runs the panel ── report (excerpt) ───────── 7-day Starter retention 5/5 critical "by the time I've acted on a dashboard it's gone" Undefined event limits 4/5 high "I can't size the bill, so I can't justify it" $$ ranked fixes ready · 47s wall · your tokens, not ours$ ▊
| Finding | Support | Severity |
|---|---|---|
| 7-day Starter retention | 5/5 | critical |
| Undefined event limits | 4/5 | high |
Excerpt — two of the run's findings. The full table, with verbatim quotes, lives in the sample report.
$$ ranked fixes ready · 47s wall · your tokens, not oursA complete verdict — ranked fixes with the strongest counter-argument — lives in the full sample report: a real, unedited panel run on a pricing page.
Airena keeps product decisions evidenced.
The persona is the survey. The report is the proof. Airena makes the two inseparable — so a decision can be argued with, not just asserted.
Research without recruiting
The panel runs when you do. Wrong answers are cheap; missing them isn't.
Evidence in every claim
Who said it, how bad it is, and what they actually said — all attached.
Runs in your host
Your Claude, Cursor, or opencode executes the panel. You pay your host, not us.
One sentence in. Every claim traceable.
One sentence in, a panel out.
Airena matches your description against a 100,000-persona subset of MatrAIx, an open research dataset of ~1M survey-derived personas, and returns the closest semantic matches. No vectors, no setup. Your panel shares your framing — the report flags where it converges instead of hiding it.
$ create_cohort("Solo SaaS Founders EU Clean")
[ok] 5 personas matched
Avery Garcia · Jordan Jones · Rowan Lee
Alex Anderson · Taylor BrownEvery claim carries its proof.
No finding ships alone: each carries who said it, how bad it is, and what they actually said. Paste the table straight into the PRD.
| Finding | Support | Severity |
|---|---|---|
| 7-day Starter retention | 5/5 | critical |
| Undefined event limits | 4/5 | high |
A verdict you can argue with.
Ranked fixes, each traceable to a finding — plus the strongest counter-argument, from the findings themselves. The report shows its work and its doubts.
FIX 1/5 Publish what “0% platform fees” excludes … COUNTER “… 0%-fee checkout plus twenty currencies the reason to move her whole shop” — Morgan Garcia (willingness 5/5)
Four ways to interrogate.
Survey
simulate_surveyPanel reactions with support counts, severity, verbatim quotes.
Web review
simulate_web_reviewStatic critique of a page, screenshot, or snapshot. No clicking.
Chat
simulate_chatFollow-up interviews that run like the first one: evidenced.
Compare
compare_versionsA reasoned critique of two variants. Not a statistical test.
Built for the ones who ship.
Builders & founders
Research without recruiting. The panel runs when you do; wrong answers are cheap, missing them isn't.
Product managers
Every claim with evidence: who said it, how bad it is, what they actually said. Paste it into the PRD.
Product marketing
Positioning, pricing, and outbound tested before launch — not after the open rates.
Marketing directors
Two variants critiqued in minutes. A fast pre-check that tells you which test to run with real users.
In their words, with receipts.
Verbatim panel speech, excerpts. Names the reports print are simulated personas, not real users.
One minute to your first panel.
Sign in, mint a key, paste one snippet. Pick your host:
{
"mcpServers": {
"airena": {
"command": "npx",
"args": ["-y", "@airena-ai/mcp@latest"],
"env": {
"AIRENA_API_KEY": "airena_live_<keyId>.<secret>",
"AIRENA_BASE_URL": "https://wooden-kookabura-444.eu-west-1.convex.site"
}
}
}
}What this is NOT.
Not real users. Personas are simulations drawn from research-grade data — they surface plausible frictions, not market truth.
Not statistically significant. 5 correlated personas, not a representative sample.
Not a replacement for user research. It's the cheap first pass that tells you which research to run — or that you can skip it.
We'd rather lose the signup than oversell the tool.