Check Your Agent

Local, evidence-backed forensics for Claude Code sessions. Claude Code writes a detailed record of every session to ~/.claude/projects. Check Your Agent turns that record into estimated API-equivalent cost, observed activity, and supported findings linked to evidence receipts. It runs entirely locally — no accounts, no telemetry, no uploads — and never modifies the original logs. Tools like ccusage tell you what you used; Check Your Agent helps you investigate why a session was expensive or unusual.

API-equivalent costs are estimates from recorded token counts and bundled pricing tables, not invoice amounts. Findings and experimental associations are leads to investigate, not causal conclusions.

Run

uvx checkyouragent

(or pipx run checkyouragent). This serves the app on port 8000, opens your browser, and uses ~/.claude/projects as the import root when it exists, otherwise ./Data. cya is an identical, shorter alias.

Flags:

• --import-root — export root to scan (default: ~/.claude/projects when present, otherwise ./Data).
• --port — bind port (default: 8000).
• --no-browser — do not open a browser.
• --demo — use the bundled synthetic demo dataset instead of a real export.

What It Shows

• Sessions — an investigation board with supported findings, observed errors, subagent fanout, event volume, duration, estimated API-equivalent cost, search, filters, and sorting.
• Session workspace — timeline, trace, subagent, findings, and raw event inspection for a single session, with receipt links from each supported finding back to the recorded events.
• Cost — estimated API-equivalent cost by project, model, and token category; cost over time, spike detection, turn distribution, and session outliers, with optional date-aware historical pricing.
• Limit hits — recorded limit notices and reconstructed five-hour windows, viewed by recorded token volume or estimated API-equivalent cost.
• Context Economics — explicitly defined opportunity signals with receipts and session drilldown; estimated and unattributed values remain distinct.
• Usage drivers — overlapping, cost- or token-weighted characteristics such as subagent activity, large input contexts, and long-running sessions. These shares are independent characteristics, not a breakdown.
• Experimental analysis — an observed-activity Usage Mindmap and subgroup associations with declared samples, uncertainty, and receipts. Neither view claims causal attribution.
• Team bundles — export allowlisted aggregates designed to omit raw conversation content at structural or team privacy levels for team Overview and Cost views. New exports use schema v3; sanitized v1 and v2 imports remain supported.
• Privacy mode — reduces visible sensitive text; screenshots still require review before sharing.

Links

• Website: https://checkyouragent.dev
• Source and full README: https://github.com/TarekAwwad/checkyouragent
• Changelog: https://github.com/TarekAwwad/checkyouragent/blob/main/CHANGELOG.md

License

Business Source License 1.1, with an Additional Use Grant: free for personal use and internal use (including commercial-internal use) by you or your organization. See the LICENSE file for the full text.