Forge your
designplan
for how people truly behave.
Not how they say they do.
15 years of design audits turned into a tool. Forge scans URLs, screenshots, and mockups across five cognitive layers and tells you what converts and what doesn't. Scored analysis, specific prescriptions, round-over-round convergence. No vibes. Measurement.
Round 3 of 4 · Score: 71 → est. 88
Reasoning
.hero-cta competes with nav for first-fixation.
User's eye path splits. Trust decision delays.
L1 violation: two equal-weight targets in 50ms window.
Fixed this round
✓ CTA contrast raised, nav weight reduced L1
✓ spacing rhythm in .feature-grid L2
✓ font-size ramp on mobile nav L2
Still open
→ .price-tag color distance too low 3.8:1
→ .cart-summary whitespace crowded L2
$ forge converge --continueHow it works
Each mode fits a different moment in your design process.
Scan
Submit a URL, screenshot, mockup, or code file. Get scored analysis across all five layers with specific findings and prescriptions.
Consult
Ask any design question. The AI reasons layer-by-layer through PFD, produces comparison tables, and references prior scans from the same session.
Synthesize
Bring in research, prior scans, and project context. Forge cross-references everything and produces optimized plans with prescriptions already applied.
Run it where you already work
Forge is an API. Call it from Claude, ChatGPT, any MCP-compatible client, or your own CLI scripts. Perception-First Design™ analysis without leaving your workflow.
$ forge analyze ./src/components/ --type directory
Scanning 14 files across 3 directories...
Foundation 92 Load within limits
L1 64 First impression needs work
L2 38 Fluency violations in nav + modal
L3 72 Bias optimization adequate
L4 88 Decision architecture strong
Score: 47/100 · Round 1 of 4
Priority: L2 Processing Fluency
→ Generating prescriptions...
→ Applying fixes to plan.md
→ Re-scanning for Round 2...The 5 Layers
Perception-First Design™ diagnostic framework
Five layers of human perception, scored individually. Fix the lowest layer first. Everything above it depends on the foundation below.
Cognitive Load Reduction
Sensory bandwidth allocation, working memory limits, and cognitive throughput.
First-Impression Architecture
The 50ms judgment. Pre-attentive pattern recognition and initial trust formation.
Processing Fluency
Typography, color relationships, spatial rhythm, and visual hierarchy effectiveness.
Perception Bias Optimization
Anchoring, framing effects, social proof, and default optimization.
Decision Architecture
Choice engineering, conversion pathway design, friction reduction, and commitment patterns.
Before Forge
"Users failed to identify the primary CTA within the first 3 seconds of exposure during attention mapping."
After 4 Rounds
"The hierarchy guides without friction. Users see the CTA first, trust the page fast, and follow the path you designed."
Investment
Choose your diagnostic level
- check 1 Analysis / Week
- check URL & Screenshot Input
- check Foundation + L1 + L2 Layers
- check Summary Report
- check PDF Export
- check Unlimited Analyses
- check All 5 Perception Layers
- check Every Input Type (URL, Image, HTML, CSS, Copy, Directory)
- check Multi-Round Convergence Loop
- check API Access & MCP Integration
- check Claude, ChatGPT & CLI Tools
- check Specific Fix Prescriptions
- check Chat with Your Analysis
- check Accumulated Learnings Database
- check Full PDF & JSON Reports
- check Everything in Professional
- check Priority Processing
- check White-label Reports
- check Multi-Client Vaults
- check Team Invites & Collaboration
- check Custom Learning Sets
Most design feedback is opinion.
This is measurement.
Five perception layers. Specific prescriptions. Round-over-round convergence tracking. Grounded in cognitive science, not vibes.
Already have an account? Sign in to Forge
55 citations · 15 years applied · Cognitive science foundation