It doesn't just guard agreements. It catches manipulation.
Every incoming claim about money, scope, or deadlines is checked against everything you've actually agreed — and the quieter plays, like 'we never agreed to that,' get caught with the receipts.
Claims, verified against your record — in seconds
A fast keyword pass clears the ~80% of email that's routine at zero cost. Anything that asserts a term gets the deep pass: the claim is extracted, your history is searched across every connected source, and a reasoning model returns a verdict — supported, contradicted, or unverifiable — with the exact sources and dates.
- Two-pass design: free keyword filter, then deep analysis only where it's warranted
- Cross-references email, calendar, and connected tools — not just one thread
- Every verdict carries its sources and a confidence score
Nine plays it watches for — in the email, and across the relationship
Some plays hide in a single message: gaslighting about what was agreed, blame quietly shifted onto you, a claim of authority someone doesn't have, a deadline redrawn after the fact. Others only show up over time — history slowly rewritten across a thread of emails, commitments talked down bit by bit, trust decaying, the balance of power tilting, priorities quietly diverging from what you both signed up for. It watches for all of them. Each alert fires only when the language pattern AND your recorded reality both hold — warnings with evidence, never paranoia.
- In-message: gaslighting, responsibility transfer, authority shift, deadline drift
- Over time: narrative drift, commitment inflation, trust decay, power shift, priority divergence
- Two-factor precision: a linguistic trigger plus contradicting history, together — it reads your stored reality only, never invents a concern
Exposure, a counter-reply, and a hard stop
For payment disputes it estimates the cash-flow exposure with its methodology shown, then drafts a correction in your voice citing the dated sources. While that flag is active, autonomous sending is frozen on the thread — so the AI can never quietly 'agree' to the new terms on your behalf.
- Directional financial exposure, methodology always shown (not legal advice)
- Counter-reply drafted in your voice, citing specific dates and sources
- Autopilot freeze on any flagged thread — enforced in code, not settings
Good to know.
Does it read every email?
A no-cost keyword pass runs on everything; only the small fraction that asserts a term, scope, or deadline goes to the deeper analysis. Personal email is excluded entirely, in code.
What if it's wrong about a risk?
Every flag shows its sources so you can judge in seconds, and a one-tap ✓/✕ on each decision trains its calibration over time. It only flags when both a language pattern and contradicting history hold, which keeps false alarms low.
Is the financial number legal advice?
No — it's a directional magnitude to help you prioritize, with the calculation shown. It's decision context, not a legal opinion.
See it on your email.
Connect Gmail and it reads your last 90 days. Your first brief shows you what it already found.
Start free — connect GmailNo credit card · Personal email permanently excluded · Export or delete everything anytime