Cost & Token Tracking
thane tracks every token Claude Code consumes — input, output, and cache reads — with per-model pricing. See costs in real time in the status bar, break down spending by workspace, and set budgets to prevent runaway spending.
Per-Model Pricing
Built-in pricing for Claude models. Costs update in real time as tokens are consumed, with accurate input/output/cache breakdowns.
Live Status Bar
Token count and cost are always visible in the status bar. Click to open the full cost breakdown panel with per-workspace and per-model details.
Per-Workspace Breakdown
See how much each workspace has spent. Identify expensive tasks, compare efficiency across runs, and allocate costs to projects.
Budget Limits
Set token or dollar limits per task or per session. thane auto-pauses Claude Code when limits are reached, giving you control over spending.
How it works
thane monitors terminal output for token usage patterns emitted by Claude Code. When Claude Code outputs token counts, thane parses them and applies the correct per-model pricing: different rates for input tokens, output tokens, and cache reads.
The cost tracking system maintains running totals at three levels: per-terminal (individual Claude Code sessions), per-workspace (all terminals in a project), and global (total across all workspaces). Each level is visible in the UI — the status bar shows the workspace total, and the cost panel shows the full breakdown.
Budget enforcement integrates with the agent queue. When you set a token limit on a queued task, the queue processor monitors the running total and pauses execution when the limit is approached. This works with the auto-pause mechanism: Claude Code gets a clean stop rather than being killed mid-operation.
Use cases
Cost allocation
Track Claude Code spending per project workspace. Export cost data for chargeback reporting or to understand which tasks consume the most tokens.
Budget management
Set daily or per-task token limits to prevent Claude Code from running up unexpected bills. Get notified when spending approaches the limit.
Prompt optimization
Compare token usage across different prompting strategies. Identify which approaches are most cost-effective for your Claude Code workflows.
Ready to try thane?
Free for personal use. Kernel-level sandboxing, split panes, embedded browser, and a 41-method API — all on your machine.