ManavThe signature layer for the agentic age, in ten scenes.
Each scene runs in under 60 seconds. Three show the signing primitive on high-risk actions. Four show the privacy-preserving policy patterns we use instead of surveillance. Three show the verification and audit surface. None of them require an account, a contract, or a download.
The signing primitive
A fresh WebAuthn ceremony bound to a specific action's canonical hash, producing an offline-verifiable receipt.
$42K vendor wire approval
A finance bot queues a wire. The release endpoint requires a fresh Manav signature bound to this exact payload. The bot can't satisfy it. Asha can.
Contract signature with human proof
A Master Services Agreement gets signed. The receipt proves a real human approved the exact contract text, with a public proof page anyone can verify.
Promote a bot to workspace owner
Promoting @opsbot to admin in Slack is a $0 action with $1M consequences. Manav gates it behind a human signature even though the API would accept the token.
Privacy-preserving policy patterns
We don't watch the worker. We require consensual signed evidence from the legitimate actor. The bad actor can't produce the signature; the action fails; no one had to be surveilled.
Step-up signing at every risk threshold
Instead of detecting mid-session worker swaps with biometric surveillance, every meaningful action demands a fresh passkey ceremony. The swap becomes useless without watching anyone.
Receipt-level analysis, not worker tracking
Detecting "same identity, two cities, one minute" via signed receipts that carry their own context with user consent. Workers loyal to one employer are invisible to the system.
Mandatory signed commits and merges
Instead of watching IDE keystrokes for "fake commits," configure the repo to reject any code on protected branches without a Manav-anchored signature. No anonymous commit can land.
Signed origin attestation
"Detecting AI-generated work" is unsolved. Instead, require a signed declaration of what tools were used. Optional process receipts cover hashes of in-progress state without surveilling content.
View and verify
The audit surface, for security teams, for workers, and for any third party who needs to verify a receipt independently.
Security team dashboard
Every signed action in one place. Filter by actor, action type, time. Replay the proof page. Export for SOC 2 / EU AI Act §14 evidence. No worker activity log, only consensual receipts.
My receipts, a worker's own proof of work
Every action you signed, exportable to your next employer as a portable verified work history. Pro-worker, not anti-worker. The receipts are yours.
Public proof page
A receipt shared with compliance, the bank, or an auditor. They verify it against Manav's published Ed25519 key with no callback to Manav required. The math works forever.
- No keystroke surveillance. No webcam telemetry. No behavioural profile.
- No biometric storage. Face ID and Touch ID stay on the user's device.
- Per-context pseudonyms, the same user presents a different key to every relying party.
- Manav is not in the verification path. A Manav outage does not break receipts.