clauditor

clauditor is a free, open-source dashboard that gives developers and teams a precise, real-time picture of what their use of Claude actually costs. It runs entirely on your own machine — reading local Claude Code session logs and your own Anthropic API usage, computing spend against current model pricing, and presenting it in a local dashboard. No account, no data leaving your device, no servers to run.

Teams building on Claude face a costly blind spot: usage is easy to rack up and hard to see. clauditor attributes every dollar to a project, model, and point in time, shows how efficiently prompt caching is being used, and flags specific, quantified opportunities to spend less — so cost control becomes a habit rather than a monthly surprise.

What it does

• Tracks spend across Claude Code sessions and your Anthropic API calls in one ledger.
• Breaks down cost by project, model, or source over any date range.
• Measures cache efficiency — the single most actionable lever on Claude cost.
• Suggests concrete savings (cheaper model, missed caching, Batch API) with real dollar estimates.
• Warns you before the bill does, with per-project and overall budgets.
• Keeps your data yours — everything runs locally.

Free and open-source under the MIT license. Runs on macOS, Linux, and Windows; requires only Python.