Synthora •

coming soon

Six phases.
Twelve agents.
One product.

Web

macOS

Windows

AI tools have sped up individual tasks, but product development remains fragmented. Research, design direction, prototyping, user testing, code, and QA live in disconnected tools – often resulting in generic outputs and lost design voice.

AI tools have sped up individual tasks, but product development remains fragmented. Research, design direction, prototyping, user testing, code, and QA live in disconnected tools – often resulting in generic outputs and lost design voice.

Synthora is a structured AI pipeline that takes your product from idea to production – with a real checkpoint before every phase moves forward. Nothing gets summarised along the way. By the final phase, the agents are still reading your full project history, not a recap of it.

PROJECT BRIEF

What are we building?

Select platform…

PHASE 01

Discovery & Research

Real personas and competitive research – the foundation every later agent reads from, not a deck no one reopens. Two specialists work in parallel: one audits your competitors and frames the core problem, the other builds personas around real goals and pain points.

3-competitor audit

2 personas + JTBD

6 Phase read through

AGENT A

DISCOVERY STRATEGIST

Competitive audit + problem framing

Audits 3 competitors, frames the core problem statement, and produces 5 "How Might We" statements plus 5–6 prioritised research questions that guide everything downstream.

AGENT B

PERSONAL RESEARCHER

PERSONA RESEARCHER

User personas + jobs-to-be-done

Builds two primary personas with goals and pain points, maps the jobs-to-be-done, and distils 3 critical design implications that carry into every phase that follows.

Phase 1 output — Discovery & Research

COMPETITIVE AUDIT

Competitor A

Onboarding

Weak

Competitor B

Mobile experience

Moderate

Competitor C

Booking flow

Strong

PRIMARY PERSONAS

Sarah, 34

Busy parent, mobile-first

Pain

Booking takes too long — loses patience after 2 taps

Goal

Confirm in under 60 seconds, no account needed

James, 41

Decision-maker, desktop

Pain

Can't see team availability at a glance

Goal

Manage recurring bookings from one dashboard

PHASE 02

IA & User Flows

The full-screen map and critical user flows – the authoritative source every later phase, including the prototype, inherits. One agent designs the navigation model end-to-end; the other walks each critical path step by step, surfacing edge cases before any screen is drawn.

10+ screens

3 critical flows

Navigation model

AGENT A

INFORMATION ARCHITECT

Product map + screen inventory

Reads the personas and research from Phase 1 to build the complete navigation model – a recommended structure, and the full-screen inventory (10+ screens) that every later phase treats as the source of truth.

AGENT B

FLOW MAPPER

Critical path flows + edge cases

Maps the 3 most important user flows step by step – happy path, error states, decision points – surfacing open UX questions before any visual design starts.

Phase 2 output — IA & User Flows

NAVIGATION MODEL

Home

└

Featured services

└

Quick re-book

Browse

└

Category filter

Bookings

CRITICAL FLOW — FIRST BOOKING

Browse

Entry from home or category

Select service

Detail view — pricing, reviews, availability

Pick date & time

No slots → suggest next available

Confirm

Guest checkout — no account required

PHASE 03

Visual Direction & Design System

Two directions are proposed: one locked – exact tokens your prototype and production code both read verbatim. A Creative Director proposes the visual language and recommends a direction, then a Design Systems Engineer turns it into literal token tables and component specs.

Full token set

20+ components

Locked & editable

AGENT A

CREATIVE DIRECTOR

Visual direction + palette

Proposes two distinct creative directions with full colour palettes and typography pairings, recommends one based on the Phase 1 personas and product context, and defines 4 product design principles.

AGENT B

DESIGN SYSTEM ENGINEER

Tokens + component specs

Takes the chosen direction and produces the literal design token tables – colour, type, spacing, radius, shadow, motion – plus specs for all 13 core components.

Phase 3 output — Design System

COLOUR TOKENS

primary

surface

text

border

muted

Display

Body text

Georgia / -apple-system

accent

#F54E00

surface

#1A1A1A

border

#2E2E2E

PRIMARY BUTTON

Book now

SECONDARY

View all

CARD RADIUS

10px

SPACING BASE

4px grid

16px / 24px / 32px

PHASE 03 - PROTOTYPE UNLOCKED

See it before you build it.

An optional, real tappable prototype generated from your locked design system feeds Phase 5 as ground truth if you build one. It distils Phases 1–3 into a compact context – your tokens, your screen list, and a set of recurring names and facts invented once – then renders one screen at a time in a sandboxed preview.

Your exact tokens

Colour, typography, and spacing read verbatim from Phase 3.

Doesn't drift

Names, prices, and facts invented once, reused everywhere.

Exportable

Download the full prototype as HTML, or any screen as PNG.

PHASE 04

Prototyping & User Testing

A full usability test plan generated automatically – synthesis strictly limited to sessions you actually ran. Agent A writes the task scenarios and screener criteria, then stops; Agent B only starts once you've pasted real observation notes from real sessions.

Task scenarios

Real findings only

Feeds Phase 5

AGENT A

TEST PLANNER

Test plan + task scenarios

Generates a prototype spec for each key flow – what to fake vs. build functionally – plus a full usability test plan: screener criteria, task scenarios with success criteria, and debrief questions.

AGENT B

RESEARCH SYNTHESISER

Findings + recommendations

Reads your raw session notes and produces themed findings, task performance analysis, and prioritised recommendations – built only from notes a person actually submitted.

Phase 4 output — Testing Plan

Complete a first booking

“You've heard about this service from a friend. Find something that works for next Saturday and book it.”

Time on task

Success rate

Error count

Find and reschedule a booking

“Something came up. Find your booking from last week and move it to Thursday.”

Navigation path

Findability

Satisfaction

Set up a recurring booking

+ 2 more tasks

Agent A stops here — run your sessions

Paste your raw observation notes. Agent B synthesises only what you observed.

PHASE 05

Production Build

A real, running scaffold in your stack – React, Swift, Flutter, Kotlin, whatever you brief it with — wired up screen by screen. No Agent A/B split here: one continuous sequence carries the largest token budget in the pipeline, from scaffold through components to every individual screen.

A real, running scaffold in your stack – React, Swift, Flutter, whatever you brief it with — wired up screen by screen. No Agent A/B split here: one continuous sequence carries the largest token budget in the pipeline, from scaffold through components to every individual screen.

Any stack

Theme from tokens

Screens wired up

AGENTS

SINGLE SEQUENCE - SCAFFOLD → COMPONENTS → PER-SCREEN

Complete production scaffold – any platform, any stack

The platform and stack are read directly from your brief. React, React Native, Next.js, Vue, Swift, Flutter, Kotlin – Synthora writes to your tech, not a default template. It produces a full project structure, a theme file with every design token, core components with all interactive states, and screens wired up from the Phase 2 inventory.

The platform and stack are read directly from your brief. React, React Native, Next.js, Vue, Swift, Flutter – Synthora writes to your tech, not a default template. It produces a full project structure, a theme file with every design token, core components with all interactive states, and screens wired up from the Phase 2 inventory.

GENERATED FILE STRUCTURE

src/theme/tokens.ts

Phase 3 tokens

TSX

src/components/Button.tsx

All states

TSX

src/components/Card.tsx

All states

TSX

src/screens/HomeScreen.tsx

Wired

tokens.ts

Button.tsx

// Generated monochrome theme tokens

export const tokens = {

color: {

primary: '#ED7A45',

surface: '#1A1A1A',

}

PHASE 6

QA Audit & Launch

A structured audit framework and a launch verdict – go/no-go built from what you actually found, nothing assumed. A QA Specialist generates the checklist and severity system, then a Launch Lead turns your real findings into a prioritised, sign-off-ready report.

WCAG AA checklist

P0–P3 severity

Go / No-go verdict

AGENT A

QA SPECIALIST

Audit framework + checklists

Generates a visual fidelity checklist against Phase 3 tokens, an interaction & animation audit framework, a full WCAG AA accessibility checklist, and a P0–P3 bug severity classification guide.

AGENT B

LAUNCH LEAD

Launch readiness report

Reads your audit findings and produces a launch verdict, findings grouped by severity, a sign-off checklist, and a retrospective on what the pipeline surfaced vs. what remained hidden until QA – citing only findings you actually submitted.

Phase 6 output — QA Audit

5 pass

1 P0

1 P2

VISUAL FIDELITY

✓

Primary colour matches token #ED7A45

Pass

Heading size incorrect on tablet breakpoint

ACCESSIBILITY

Button contrast ratio 2.8:1 — fails WCAG AA

INTERACTIONS

✓

Booking flow completes in 4 taps

Conditional go — 1 P0 to resolve

Fix contrast on primary button, then cleared for launch

CHECKPOINTS

Three kinds of checkpoints – never a black box

Not every gate works the same way. Synthora uses whichever fits what's actually at stake in that phase – but every phase ends with a human in the loop before anything moves forward.

STANDARD GATE

Review, then decide

Both agents run automatically end to end, then you review the finished result. Used for Phases 1, 2, and 6's final verdict.

Approve – locks it in, starts the next phase

Refine – add guidance, rerun this phase

Redo – wipe and rerun from scratch

LOCK & EDIT GATE

Edit the exact wording

The combined output lands directly in an editable text box – you can rewrite it by hand. Saving finalises the phase immediately, no separate approval click.

Used where exact wording matters most

Phase 3 (design system) and Phase 5 (production build)

What you save is exactly what downstream phases read

HUMAN INPUT GATE

Real people, real findings

Agent A stops and waits – a real person must submit real content before Agent B is allowed to run at all.

Used for Phases 4 and 6

Agent B may only cite what was actually submitted, never invent

Ends in a standard gate once Agent B finishes

Synthora

Start with a brief.
End with a product.

One brief describing your product sets twelve specialised agents in motion – from user research to production code. Every decision you make along the way shapes what the next agent reads.

Six phases.Twelve agents.One product.

Six phases.Twelve agents.One product.

Six phases.Twelve agents.One product.

AI tools have sped up individual tasks, but product development remains fragmented. Research, design direction, prototyping, user testing, code, and QA live in disconnected tools – often resulting in generic outputs and lost design voice.

Synthora is a structured AI pipeline that takes your product from idea to production – with a real checkpoint before every phase moves forward. Nothing gets summarised along the way. By the final phase, the agents are still reading your full project history, not a recap of it.

Discovery & Research

Real personas and competitive research – the foundation every later agent reads from, not a deck no one reopens. Two specialists work in parallel: one audits your competitors and frames the core problem, the other builds personas around real goals and pain points.

DISCOVERY STRATEGIST

Competitive audit + problem framing

PERSONAL RESEARCHER

PERSONA RESEARCHER

User personas + jobs-to-be-done

IA & User Flows

The full-screen map and critical user flows – the authoritative source every later phase, including the prototype, inherits. One agent designs the navigation model end-to-end; the other walks each critical path step by step, surfacing edge cases before any screen is drawn.

INFORMATION ARCHITECT

Product map + screen inventory

FLOW MAPPER

Critical path flows + edge cases

Visual Direction & Design System

Two directions are proposed: one locked – exact tokens your prototype and production code both read verbatim. A Creative Director proposes the visual language and recommends a direction, then a Design Systems Engineer turns it into literal token tables and component specs.

CREATIVE DIRECTOR

Visual direction + palette

DESIGN SYSTEM ENGINEER

Tokens + component specs

See it before you build it.

Your exact tokens

Doesn't drift

Exportable

Prototyping & User Testing

A full usability test plan generated automatically – synthesis strictly limited to sessions you actually ran. Agent A writes the task scenarios and screener criteria, then stops; Agent B only starts once you've pasted real observation notes from real sessions.

Test plan + task scenarios

Findings + recommendations

Production Build

A real, running scaffold in your stack – React, Swift, Flutter, Kotlin, whatever you brief it with — wired up screen by screen. No Agent A/B split here: one continuous sequence carries the largest token budget in the pipeline, from scaffold through components to every individual screen.

A real, running scaffold in your stack – React, Swift, Flutter, whatever you brief it with — wired up screen by screen. No Agent A/B split here: one continuous sequence carries the largest token budget in the pipeline, from scaffold through components to every individual screen.

Complete production scaffold – any platform, any stack

QA Audit & Launch

A structured audit framework and a launch verdict – go/no-go built from what you actually found, nothing assumed. A QA Specialist generates the checklist and severity system, then a Launch Lead turns your real findings into a prioritised, sign-off-ready report.

Audit framework + checklists

Launch readiness report

Three kinds of checkpoints – never a black box

Not every gate works the same way. Synthora uses whichever fits what's actually at stake in that phase – but every phase ends with a human in the loop before anything moves forward.

Review, then decide

Edit the exact wording

Real people, real findings

Start with a brief.End with a product.

One brief describing your product sets twelve specialised agents in motion – from user research to production code. Every decision you make along the way shapes what the next agent reads.

Coming soon…

Six phases.
Twelve agents.
One product.

Six phases.
Twelve agents.
One product.

Six phases.
Twelve agents.
One product.

Start with a brief.
End with a product.