Uxia (uxia.app) platform help — an AI/synthetic-tester usability-testing tool: AI "testers" modeled on demographic/behavioral profiles explore a design, prototype, or live URL and surface UX friction in minutes, returning transcripts, heatmaps, SUS/SUPR-Q scores, accessibility checks, and reports across five test types. Figma/Jira/Notion/Linear integrations plus a marketed-but-undocumented MCP link to Claude/ChatGPT/Gemini; no public REST API. Use when setting up an AI user test, testing a Figma prototype or live site with synthetic testers, reading a SUS/SUPR-Q or accessibility result, wondering how realistic AI testers are or whether they replace real users, hitting the free-trial credit limit, or exporting results without an API. Do NOT use for comparing usability/research/validator tools or the validate-before-building method (use /sales-idea-validation), real-human usability testing (use /sales-userbrain or /sales-uxtweak), or synthetic-persona interviews (use /sales-syntheticusers).
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Une commande directe contourne le prompt de vérification. Examinez la source avant de l'exécuter.
Uxia (uxia.app) platform help — an AI/synthetic-tester usability-testing tool: AI "testers" modeled on demographic/behavioral profiles explore a design, prototype, or live URL and surface UX friction in minutes, returning transcripts, heatmaps, SUS/SUPR-Q scores, accessibility checks, and reports across five test types. Figma/Jira/Notion/Linear integrations plus a marketed-but-undocumented MCP link to Claude/ChatGPT/Gemini; no public REST API. Use when setting up an AI user test, testing a Figma prototype or live site with synthetic testers, reading a SUS/SUPR-Q or accessibility result, wondering how realistic AI testers are or whether they replace real users, hitting the free-trial credit limit, or exporting results without an API. Do NOT use for comparing usability/research/validator tools or the validate-before-building method (use /sales-idea-validation), real-human usability testing (use /sales-userbrain or /sales-uxtweak), or synthetic-persona interviews (use /sales-syntheticusers).
argument-hint
[describe what you need help with in Uxia]
license
MIT
version
1.0.1
tags
["sales","pre-launch","platform"]
Uxia Platform Help
Uxia (uxia.app) is an AI / synthetic-tester usability-testing tool — the synthetic counterpart to
the real-human usability tools (Userbrain, UXtweak). Instead of recruiting people, Uxia runs AI "testers"
(LLM agents modeled on demographic + behavioral profiles + any uploaded persona docs) that explore your
design, prototype, or live URL, "think aloud," and surface UX friction in ~5 min → transcripts,
heatmaps, SUS / SUPR-Q usability scores, accessibility checks, and actionable reports across five test
types: AI User Test, AI Live Test (paste a live URL), AI User Research, Accessibility Test,
and a Human Test (real humans — enterprise/custom-plan only).
Two things to say almost every time (both are Step 4 imperatives):
Synthetic ≠ real ≠ demand. AI testers are great for early iteration (stress-test flows 20× a
sprint) but don't replace real humans (the founder says so), and a smooth run does not prove
anyone will pay — validate the final version with real testers (/sales-userbrain, /sales-uxtweak)
and take a build-or-not go/no-go to a real behavior test (/sales-idea-validation).
No public REST API — only a marketed-but-undocumented MCP link (Claude/ChatGPT/Gemini; no docs/
endpoints/schemas). Any "pipe results into my CRM/warehouse/Slack" ask is manual export or the MCP
link — discover MCP tools at runtime, don't invent endpoints; for a documented pipeline route to
/sales-userintuition (or /sales-syntheticusers / /sales-ditto).
Step 1 — Gather context
If references/learnings.md exists, read it first for accumulated platform knowledge.
Ask only what you can't infer:
What do you want from Uxia?
A) Set up a test — a test type (AI User Test / Live Test / User Research / Accessibility), tasks, Figma prototype or live URL
B) Configure testers — define an audience (demographics/tech literacy), upload persona docs, set tester count
C) Read results — transcripts, heatmaps, SUS/SUPR-Q scores, the report
D) Judge validity — how realistic the synthetic testers are, whether they replace real users
E) Get data out / automate — export a report or wire up the MCP link (no REST API)
F) Choose — Uxia (synthetic) vs Userbrain/UXtweak (real humans) vs Synthetic Users/Ditto (synthetic interviews)
Are you testing something you've already built/designed, or still deciding whether to build? The
second is an idea-validation question, not a usability question — flag it in Step 2.
Skip-ahead: if the user's prompt already names the task and their question is specific, go to Step 3.
Step 2 — Route or answer directly
If the user's question is about…
Route to
Real-human think-aloud usability testing (Uxia's testers are synthetic)
/sales-userbrain {question}
Card sorting / tree testing / first-click IA tests, or the broad usability suite
/sales-uxtweak {question}
Comparing usability/research/validator tools across the market, or the validate-before-building method
/sales-idea-validation {question}
Synthetic interviews/surveys against AI personas (a different job from usability testing)
/sales-syntheticusers or /sales-ditto{question}
An API / webhook-native research or interview pipeline (Uxia has no public API)
/sales-userintuition {question}
Running a real behavior demand test (smoke-test page, waitlist, pre-sale)
/sales-idea-validation or /sales-funnel{question}
When routing, give the exact command: "This is a {domain} question — run: /sales-userbrain {original question}"
Otherwise, answer Uxia-specific questions using Step 3.
Step 3 — Uxia platform reference
Read references/platform-guide.md for the full reference — the capability/automation-surface table
(five test types, each UI-only / MCP-only), the synthetic-tester data model, best-effort pricing and plan
gates (free-trial credits, the flat-monthly-vs-credit ambiguity, the enterprise-gated Human Test + SSO/SCIM),
the SUS/SUPR-Q frameworks, the Figma/Jira/Notion/Linear integrations, and the no-API / MCP data-out playbook.
Uxia has no public REST API, so references/uxia-api-reference.md documents the (undocumented) MCP
surface and integrations, not endpoints.
Answer using only the relevant section — don't dump the full reference.
Step 4 — Actionable guidance
You no longer need the platform guide — focus on the user's specific situation.
Say the caveat: synthetic testers are an AI opinion, not real behavior, and not demand. Whatever the
user asks, make explicit that Uxia's testers are AI, not real people — great for early-stage
iteration (find friction, test flows many times cheaply) but not a replacement for real humans
(they lack genuine intuition/emotion/randomness), and a clean synthetic run does not prove anyone
will pay. Recommend the pattern the founder describes: iterate with synthetic testers, then
validate the final version with real testers (/sales-userbrain, /sales-uxtweak). If the real
question is build-or-not, take the go/no-go from a real behavior test via /sales-idea-validation.
Flag the automation reality: no public REST API, MCP marketed but undocumented. When automation or
export comes up, state plainly there is no public REST API and no webhooks — the only programmatic
surface is a marketed MCP link (ChatGPT/Claude/Gemini) with no published docs, endpoints, or
schemas. Tell the user to discover the MCP tools at runtime and not to invent endpoints; "sync to
HubSpot/Snowflake/Slack" is a manual export otherwise. For a documented pipeline, route to
/sales-userintuition (real-human, REST API + webhooks + MCP) or /sales-syntheticusers / /sales-ditto
(synthetic studies with a real API).
Read the SUS/SUPR-Q scores as directional, not absolute. Uxia scores designs with the standard
System Usability Scale (SUS) and SUPR-Q frameworks — useful for ranking variants against each
other, but because the raters are AI, treat the absolute number as directional and trust relative
comparisons (variant A vs B) over a single score. Pair the score with the qualitative transcripts and
heatmaps — the why the testers struggled — which is the more actionable output.
Fix shallow feedback with better tasks, not more testers. If the AI testers "narrate the screen" or
say "looks easy," that's a task-design problem — write goal-based scenario tasks with no hints
("You want to cancel your subscription — do it"), add open follow-up questions, and define a
specific audience (demographics + uploaded persona docs) so a "novice" behaves differently from an
"expert." A handful of testers per audience surfaces most friction; more mostly repeat it.
Present all pricing as best-effort and point to uxia.app. Pricing is inconsistent across sources
— the site shows a Free trial (~50 credits, 1 audience, ~5 testers/test) plus a Custom/Enterprise
tier (unlimited, audience enrichment, Human Test, SSO/SCIM), while third-party listings describe a
flat monthly Basic/Pro model ("unlimited tests, no 6-figure contracts"). Say the number is
best-effort, note the Human Test and enterprise security are custom-plan-only, and point to
uxia.app/pricing to confirm.
If you discover a gotcha or tip not in references/learnings.md, append it there with today's date.
Gotchas
Best-effort from research (2026-07) — pricing, plan gates, the five test types, SUS/SUPR-Q scoring, and the (undocumented) MCP/integration surface move; verify at uxia.app.
Synthetic ≠ real ≠ demand. Uxia's testers are AI, not humans. Use them for fast early iteration,
but they don't replace real users (the founder agrees) and a clean run is not proof anyone will
buy — validate the final design with real testers (/sales-userbrain//sales-uxtweak) and take a
build-or-not go/no-go to a real behavior test (/sales-idea-validation).
It's "not just an LLM," but it's still an LLM under the hood. Uxia layers personas, task simulation,
consistency rules, and aggregation on an LLM — better than raw ChatGPT, but still can be confidently wrong and skews agreeable/generic.
No public REST API; MCP is marketed but undocumented. The only automation surface is an MCP link to
Claude/ChatGPT/Gemini with no published endpoints/schemas — don't invent them; discover at runtime.
Everything else is manual export. Route a documented pipeline to /sales-userintuition.
The Human Test is enterprise-only (as are audience enrichment, SSO/SCIM) — the free/self-serve tier is synthetic testers only.
Pricing is ambiguous — marketing says "flat pricing, unlimited tests"; the site shows a credit-limited
free trial + Custom. Treat every figure as best-effort and confirm at uxia.app/pricing.
SUS/SUPR-Q from AI raters is directional. The scores use real usability frameworks but the raters are
synthetic — read them as a relative signal between variants, backed by the transcripts/heatmaps, not
an absolute usability verdict.
Related skills
/sales-userbrain — The real-human video-first usability peer — the pick to validate a final design with actual testers after iterating with Uxia's synthetic ones. Install: npx skills add sales-skills/sales --skill sales-userbrain -a claude-code
/sales-uxtweak — The real-human usability + information-architecture suite (card sort, tree test, first-click) — route here for IA methods Uxia doesn't do, or to test with real users. Install: npx skills add sales-skills/sales --skill sales-uxtweak -a claude-code
/sales-syntheticusers — The synthetic-research peer (AI-moderated interviews/surveys vs personas, with a real REST API) — a different job from Uxia's task-based usability testing. Its synthetic message/creative-tester sibling is /sales-evelance (AI personas score copy/creative/product on 12 psychology metrics) — reach for Evelance when the question is "does my message land", Uxia when it's "is my design usable". Install: npx skills add sales-skills/sales --skill sales-syntheticusers -a claude-code
/sales-idea-validation — The tool-agnostic validate-before-building method + the full research/usability/validator tool landscape (use this to decide build-or-not; a passing synthetic usability test is not demand). Install: npx skills add sales-skills/sales --skill sales-idea-validation -a claude-code
/sales-userintuition — The API/webhook/MCP-native real-human interview peer — the pick when you need a documented, programmatic research pipeline Uxia can't offer. Install: npx skills add sales-skills/sales --skill sales-userintuition -a claude-code
/sales-funnel — Build the smoke-test / fake-door landing page that turns a design hypothesis into a real demand test. Install: npx skills add sales-skills/sales --skill sales-funnel -a claude-code
/sales-do — Not sure which skill to use? The router matches any sales objective to the right skill. Install: npx skills add sales-skills/sales --skill sales-do -a claude-code
Examples
Example 1: "Are these AI testers real enough to trust — do they replace real user testing?"
User says: "Uxia gives me feedback in 5 minutes from AI testers. Can I just use this instead of recruiting real users?"
Skill does: Explains Uxia's testers are synthetic (LLM agents on demographic/behavioral profiles),
not people — useful for fast early iteration but not a replacement for real humans (they lack real
intuition/emotion/randomness, skew agreeable), and a smooth synthetic run is not proof anyone will pay.
Recommends the founder's pattern — iterate with synthetic testers, then validate the final version with
real testers (/sales-userbrain, /sales-uxtweak) — and, if the question is build-or-not, taking the
go/no-go from a real behavior test via /sales-idea-validation.
Result: The user keeps Uxia for cheap iteration but plans a real-human validation round before shipping.
Example 2: "How do I get every finished Uxia test into our data warehouse / Slack automatically?" (developer/automation)
User says: "I want each completed test's transcript and report to flow into Snowflake and post a summary to Slack."
Skill does: States plainly Uxia has no public REST API and no webhooks — the only programmatic
surface is a marketed MCP link to Claude/ChatGPT/Gemini with no published endpoints or schemas, so
you must discover the MCP tools at runtime and not invent endpoints; absent MCP access, data-out is
manual export of the report/transcript to run your own ETL. If a documented live pipeline is required,
routes to /sales-userintuition (REST API + HMAC webhooks + MCP) or /sales-syntheticusers /
/sales-ditto (synthetic studies with a real API), noting those are interview/survey studies, not Uxia's
task-based usability tests.
Result: The user stops hunting for a REST API that doesn't exist and either uses the MCP link or a manual export routine.
Example 3: "What does the free plan give me, and how does Uxia score my design?"
User says: "I'm a solo maker. What can I do on Uxia's free tier, and what's this SUS/SUPR-Q number it shows?"
Skill does: Gives best-effort free-tier limits (a credit-limited free trial — ~50 credits, 1
audience, ~5 testers/test — with the Human Test, audience enrichment, and SSO/SCIM gated to
Custom/Enterprise), flags that pricing is inconsistent (flat-monthly marketing vs on-site credits)
and points to uxia.app/pricing. Explains the scores use the standard SUS and SUPR-Q frameworks
but, because the raters are AI, to read them as a relative signal between design variants (backed by
the transcripts and heatmaps), not an absolute usability verdict.
Result: The maker runs a free-trial test, compares two variants by relative SUS, and reads the transcripts for the why.
Troubleshooting
"The AI testers just narrate the screen / say it's easy — no real insight"
Symptom: Synthetic testers describe the page or say "looks easy to use," with nothing actionable.
Cause: Vague, leading, or hint-laden tasks, or an undefined audience, so the AI has nothing real to do.
Solution: Write goal-based scenario tasks with no hints ("You want to return an item you bought
last week — do it"), add open follow-up questions ("what did you expect?", "what was confusing?"), and
define a specific audience (demographics + tech literacy + uploaded persona docs) so a "novice" behaves
differently from an "expert." Read the transcripts/heatmaps for the why, and remember even a clean
result proves the design works in simulation, not that anyone will pay.
"I can't find Uxia's API to automate this"
Symptom: The user is looking for REST endpoints, webhooks, or a Zapier app and can't find them.
Cause: Uxia has no public REST API and no webhooks — the only automation surface is a marketed
but undocumented MCP connection to Claude/ChatGPT/Gemini (no published endpoints/schemas).
Solution: Don't invent endpoints — if you have MCP access, discover the tools at runtime; otherwise
data-out is manual export of the report/transcript. For a documented, API-native pipeline, use
/sales-userintuition or /sales-syntheticusers / /sales-ditto (synthetic studies with a real API).
"Can I test with real humans, and why is my score different from a real test?"
Symptom: The user wants real-human testing on Uxia, or distrusts the AI SUS/SUPR-Q number.
Cause: Uxia's default testers are synthetic; the Human Test is gated to the Custom/Enterprise
plan, and AI raters produce a directional, not absolute, usability score.
Solution: On self-serve, expect synthetic testers only — for real humans upgrade to the enterprise
Human Test or run a real-human tool (/sales-userbrain, /sales-uxtweak). Treat the AI SUS/SUPR-Q as a
relative signal between variants (trust A-vs-B ranking over the absolute number), backed by transcripts.
