SignOff is
A collective inbox for multiple agents to reach to get human input

I hand work to several AI agents. They keep stopping to ask me something, while I’m away.
Swap the hero photo?
Prototype the new nav?
Send the meeting recap?
Mock up the onboarding flow?
Should I plan the weekend trip?
Book movie tickets?
Archive the old project?
Cook in or order dinner?
Move the 3pm meeting to 4?
New colour for the footer?
Publish the SignOff page?
Driving lesson at 9 or 11?
Split the project into two?Work sits blocked
If I miss a question while I'm away, it will sit for hours.
Claude Code✻ Waiting for input…
2h 17m
Every ask is different
Questions can range from text-based to multi-option selects.
Cook in tonight or order?
Driving lesson?I lose track
Requests pile up quickly when using 5+ agents.
6
9
4SignOff collects all requests in one place so decisions are made faster and agents are not blocked on approvals.
Batched requests to reduce token usage on concurrent tasks
Approve, deny or discuss by sending notes to your agent
Cards designed to match different ways an agent asks questions
Before
portfolio — claude
⏺ Bash(vercel deploy --prod)
⎿ Waiting…
Bash command
vercel deploy --prod
Deploy the portfolio to production
Do you want to proceed?
❯ 1. Yes
2. Yes, and don't ask again for vercel deploy commands
3. No, and tell Claude what to do differently (esc)
Waiting for you · 3h 12m
After

Tasks with deadlines communicate when they expire
Push notifications and badges to bring attention to open items
The app uses MCP calls to communicate with all agents, and translates their text based questions into UI to interact with.






Waiting for a call…
ask_human({ from: "Dr Blob", kind: "approval", title: "Book movie tickets?" })
pending
{ status: "answered", answer: }
Dr BlobBook movie tickets?ApproveApproved ✓DenyI use it every day. The first real answer went from my phone into the very session that was building SignOff.
4 min median wait time
per request, down from ~17 min
368 requests answered
in the first 14 days
~1.1M tokens saved
by asking in lists: 54% fewer answer checks


