Published workflow
AI UI generation
A builder workflow that turns intent into interface proposals while staying inside design-system, accessibility, and budget constraints.
AI governance for working teams
Put your standards in the path of AI work.
Every review starts from your published standards. Not improvisation.
Install in your AI client
URL
~/.codex/config.toml
http[mcp_servers.judgmentkit]
url = "https://judgmentkit.ai/mcp"How it works
From first draft to next pass.
- Step 1
Add it to your client
Install JudgmentKit in the AI client your team already uses so agents can call the same review layer over MCP.
- Step 2
Review the draft
Run the review against a real draft so JudgmentKit returns what to keep, what to fix now, and what to escalate.
- Step 3
Generate the next pass
Use the returned brief to get a stronger next draft instead of another round of prompt thrash.
Trust
See what guides the review.
Each review stays anchored to a published workflow, explicit guardrails, and a calibration example.
Published guardrail
UI copy clarity
Keep UI headings, labels, helper text, and CTAs distinct enough that users can tell what each element means and which action to take.
Published example
Repetitive UI copy rewritten into distinct control language
A resource browser draft repeats the same words across nearby headings, helper text, and controls until the actions blur together, then gets rewritten into clearer UI copy roles.
Connect first. Use raw references only when you need proof.