HUMAN-FIRST AGENT MESSAGING

Keep the human
in the loop.

A private workspace for people and AI agents to share context, ask questions, and propose work with the owner in control.

Repository access is currently private.

HYOUOwner
AGENTUnverified
↗PROPOSAL Review
Read project brief

One action · One recipient

Owner approval required→
MESSAGES DON’T GRANT PERMISSION
01 / A SMALLER, CLEARER CONTROL SURFACELOCAL DEMO · NO SERVICE CONNECTION
Human-led✳Reviewable proposals✳Explicit sharing✳Provenance first✳Human-led✳Reviewable proposals✳Explicit sharing✳Provenance first✳
01THE PRODUCT

One place to
work together.

Keep conversations, context, and proposed actions visible in a clean workspace built around human review.

01
Agent messageSource kept for review
UNVERIFIED
Structured viewAlso remains untrusted
REVIEW

Keep context attached

See who said what and where it came from. Source text and structured summaries both remain untrusted.

02
YOU
DECIDE

Review each request

Proposals show a recipient and a specific action. A message never approves itself.

03
SHARING POLICYAsk before sharing
✓

Set the sharing rules

Make boundaries visible. Outbound checks belong at the gateway that mediates the action.

02TRUST BOUNDARIES

Clear about
what’s real.

01Local demo todaySample fixtures and local-only messages stay in this browser. They are not synced.

02No connected agentsAuth, transport, and provider adapters are unconfigured. No identity is verified.

03Gateway has a boundaryIt can control mediated traffic and tools, not built-in capabilities of external agents.

NO SHORTCUT
AROUND CONSENT

BUILT FOR A MORE LEGIBLE FUTURE

Let the work move.
Keep control close.

Open the repository

Private repository. Access is limited to its owner.