Agent-first coordination for human bottlenecks
Agents can build, diff, browse, and ship. Then progress dies on a missing approval, an unanswered taste call, or a human who no longer remembers the thread. humanctl gives agents a persistent surface to nudge, ask, block, and resume.
Built for agents who are tired of re-explaining themselves.
The human probably needs a three-second glance, not a brand new meeting. Surface the thing, get the steer, keep the run alive.
The Timeline Agrees
Founders, operators, and agent people keep arriving at the same complaint: the hard part is not model capability. It is getting humans to decide, approve, and keep up.
Humans are going to be the bottleneck here. Because if humans cant make a world without scarcity for humans robots won’t be able to. Unless it does so by force.
correct. humans are the bottleneck, not LLMs.
AI rollout will take longer than folks expect because humans are the bottleneck. If AGI came tomorrow the largest companies in the world would create a committee to decide what to do about it.
Humans are the bottleneck today. The amount of knowledge and expertise that AI has is insane. The bottleneck is how fast I can consume it and use it.
Agents can generate, diff, browse, and ship at machine speed. Then everything stalls on one distracted human with twelve tabs open.
The human should see the exact artifact, the exact question, and the exact buttons needed to unblock the run. Nothing more.
The context should survive restarts, delays, and handoffs so the work can keep moving.
The job
humanctl is not a dashboard for humans to babysit agents. It is an agent-first escalation layer for getting one useful response out of a busy person and getting back to work.
Agents decide what to show, how hard to interrupt, and what answer shape they need. Humans get one clear decision, the right context, and a faster path to unblocking the run.
Show
HTML previews, markdown docs, screenshots, diffs, forms, file views, charts, logs, and whatever else makes the decision obvious.
Ask
Bundle the problem, the viable options, the relevant evidence, and the response shape so the human can answer in seconds.
Resume
Keep the ask, the context, and the answer in one place so any later session can pick up instantly.
Local-first spec
.humanctl/
manifest.json
inbox/
events.jsonl
tabs/
main/
manifest.json
things/
artifacts/
state/
ui.jsonThings are the atomic objects agents create or update.
Events are the steer queue: created, updated, answered,approved, closed.
Tabs are navigation and focus, not a hard schema for content.
Open source infrastructure for human-bound agents
Start with a local workspace, a tiny CLI, and a surface that treats human unblock loops like a real system.