Session cost
Estimate list-price spend from the usage records attached to Claude responses.
Private browser dashboard · JSONL stays local
Load one or more Claude Code sessions to see estimated session cost, token and model usage, tool activity, aggregate spend and cost per session. No spreadsheet and no raw JSON counting required.
Local cost dashboard
1 session loaded
Estimated spend
$0.02
1/1 sessions priced
Token usage
30.6K
461 output
Cost per session
$0.02
average list-price estimate
Tool activity
3
tool calls across sessions
Cost over time
8/29/2026, 9:00:00 AM · 5.2s
Token mix
Model usage
| Model | Sessions | Input + cache | Output | Estimated cost |
|---|---|---|---|---|
| claude-sonnet-4-6 | 1 | 30.1K | 461 | $0.02 |
Files stay in this tab. Estimates use standard Anthropic API list prices encoded in tracehouse; subscriptions, cloud-provider pricing, fast mode, data residency, batch discounts and negotiated rates can differ.
Estimate list-price spend from the usage records attached to Claude responses.
Separate uncached input, cache writes, cache reads and model output.
Compare sessions chronologically and spot expensive outliers.
Count the tool calls behind the token and cost total.
When a Claude Code transcript contains response usage, the tracker prices uncached input, output, five-minute cache writes and cache reads for recognized Claude models. Unknown model versions remain unpriced instead of receiving a fabricated estimate.
Token usage is taken from assistant-message usage objects in the session transcript.
Costs are estimates based on standard Anthropic API list prices, not a billing statement.
Claude Pro, Max and Team subscriptions do not necessarily map to per-token API pricing.
Cloud providers, data residency, fast mode, prompt caching rules and negotiated rates may change the amount billed.
Verify current Anthropic pricingWant usage recorded automatically?
Automatic Claude Code monitoring connects cost and token usage to the prompts, tool calls, outputs, timing and errors inside each session.