Blocked session
A permission prompt sits idle in some hidden tab and you miss it for ten minutes.
Mission control for Claude Code
Know which agent is blocked, burning budget, waiting for approval, or stalled, and intervene without tab hunting.
Why it exists
A permission prompt sits idle in some hidden tab and you miss it for ten minutes.
One agent keeps spending while doing low-value work and you only notice after the fact.
A session looks alive but is actually stalled, idle, or missing transcript telemetry.
You waste time switching tabs just to understand state before you can intervene.
What claudectl does
Status, burn rate, context usage, activity, CPU, memory, and subagent rows in one place.
Approve prompts, send input, jump to the right tab, or kill a runaway session from one pane.
Set budgets, alert at thresholds, and auto-kill over-budget runs before they drift.
Run dependency-ordered task graphs, keep logs per attempt, and resume where it matters.
Trigger desktop notifications, shell hooks, and webhooks when sessions need attention.
Turn sessions and dashboards into GIFs and casts you can use in docs, posts, and demos.
Core workflows
Find every session waiting for approval and clear it immediately.
claudectl --watch --focus attention
Track live burn rate, set budgets, and auto-kill over-budget runs.
claudectl --budget 5 --kill-on-budget
Run multiple Claude Code sessions without losing operational control.
claudectl --run tasks.json --parallel
Install
claudectl --demo
claudectl --demo --record demo.gif
Use demo mode for screenshots, evaluation, and zero-risk recording.
brew install mercurialsolo/tap/claudectl
cargo install claudectl
Use plain claudectl once you already have a live Claude Code session running.
Terminal coverage
Run claudectl --doctor to verify launch, switch, input, and approve support in your current terminal.
Launch
This site is designed to answer the top-of-funnel question fast: why someone running multiple Claude Code sessions should care.