Agent Queue
Queue up multiple Claude Code tasks for headless execution. thane processes them in priority order, auto-pauses when token limits are reached, and lets you review results before moving to the next task.
Headless Execution
Submit tasks to the queue and let thane run them via Claude Code's headless mode (claude --print). Each task runs unattended in its own working directory — no babysitting required.
Priority-Based Scheduling
Assign priority levels to queued tasks. Critical bug fixes run before documentation updates. Reorder the queue at any time.
Token Limit Auto-Pause
Set a token budget per task or per session. thane automatically pauses Claude Code when the limit is reached, preventing runaway costs.
Status Monitoring
The agent queue panel shows running, queued, completed, and failed tasks with live status. Click any task to see its terminal output.
Task History & Output Logs
Completed and failed tasks are preserved in a processed tasks panel. Click any entry to view its full output log — review what Claude Code did, even after the task finishes.
How it works
The agent queue is a priority-ordered list of tasks, each with a prompt, target workspace, and optional configuration (token limit, priority, timeout). When you submit a task, it enters the queue with status 'queued'. The queue processor picks up the highest-priority task, creates a terminal in the target workspace, and sends the prompt.
While a task is running, thane monitors Claude Code's terminal output for completion signals. The status bar shows the active task, queue depth, and estimated token usage. You can cancel running tasks, reorder queued tasks, or add new tasks at any time through the queue panel or the CLI.
Token limit enforcement works by tracking cumulative token usage per task. When a task's usage approaches its limit, thane sends a pause signal and marks the task as paused. You can review Claude Code's progress, increase the limit, or cancel the task. This prevents a single runaway task from consuming your entire API budget.
Use cases
Batch code review
Queue up review tasks for every PR in your backlog. thane works through them overnight, and you review the results in the morning.
Automated refactoring
Submit a list of refactoring tasks with priorities: 'Migrate module A to new API', 'Update tests for module B', 'Fix deprecation warnings'. thane processes them in order.
Multi-step Claude Code workflows
Chain tasks where each depends on the previous one's output. Run tests after code changes, update docs after API changes, deploy after tests pass.
Ready to try thane?
Free for personal use. OS-level sandboxing, split panes, embedded browser, and real-time audit trail — all on your machine.