A clearer yes.
A kinder not yet.
Compare the original brief with the next “small change.” Find evidence, surface uncertainty, and start a better conversation.
Your local AI workspace
AI not loaded · analysis unavailable
*Load AI contacts jsDelivr and Hugging Face to download open-source code and model assets. Primary assets are about 46 MB decoded, plus small configs; actual transfer and browser resources vary. After loading, text embeddings run in your browser; this app does not send brief or request text to a server. Provider requests expose normal network metadata. Browser caching is best effort, not an offline guarantee. A fresh page needs network access to load the AI code.
Two inputs. One calmer conversation.
Not a verdict. A place to start.
Experimental, uncalibrated thresholds: ≥0.72 possible coverage; 0.42–0.72 ambiguous; lower scores potential changes. Scores are cosine similarity, not probabilities. Negation, quantities, dates, pricing and dependencies can defeat semantic matching. Exclusion matches are lexical candidates, not legal conclusions.
A polite first draft. Yours to edit.
Evidence over confidence
Every result shows matching brief passages. Read them in context and compare with the full agreement.
Keep humans in the loop
Nothing is sent, priced, approved or signed. You choose the next step with your client.
Built for a real boundary
A small open-source prototype for freelance web developers. Synthetic examples only; no customer claims.