Your data never leaves your browser. All processing happens locally on your device.
Important: What's in these files
Claude Code JSONL files contain your full conversation history including prompts, code, file contents, and command outputs. This tool only reads token counts and costs, but the raw files are loaded into browser memory during processing. For maximum safety, use the MCP method below which strips sensitive data first.
Recommended: Use the MCP tool (strips sensitive data first)
- 1
Install: claude mcp add claude-usage -- python3 <(curl -s https://findutils.com/mcp/claude-usage-mcp.py) - 2Ask Claude: "analyze my usage"
- 3Drag /tmp/claude-usage-clean.json here for charts
Or drop raw .jsonl files directly (all data stays in browser)
Drop your Claude Code session files or click to browse
Accepts .jsonl files from ~/.claude/projects/ or a pre-stripped .json from the MCP tool
~/.claude/projects/*/*.jsonl
How to Analyze Your Claude Code Usage
- 1
Open the Analyzer
Navigate to the Claude Code Usage Analyzer tool page. - 2
Locate Your Session Files
Open your file manager and navigate to ~/.claude/projects/. You'll see subdirectories for each project with .jsonl files inside. - 3
Drop Files
Drag the folder or individual .jsonl files onto the drop zone. The analyzer will process them instantly in your browser. - 4
Explore Your Dashboard
View cost trends, token usage, session counts, model breakdown, and cache analysis. Use the time range and model filters to focus on specific periods or models. - 5
Export Your Data
Click JSON or CSV to download a summary of your usage data for spreadsheets or further analysis.
Why Use Claude Code Usage Analyzer?
Claude Code subscriptions show a progress bar with your usage percentage, but no breakdown by model, project, or day. This tool reads the JSONL session files that Claude Code already writes to your machine at ~/.claude/projects/ and turns them into a visual dashboard.
Every session, every turn, every token is logged locally: input tokens, output tokens, cache reads, cache creation, model name, and timestamp. This analyzer processes those files entirely in your browser to give you cost estimates, usage trends, and cache analysis without sending a single byte to any server.
Whether you want to understand which model costs the most, track your daily usage patterns, or investigate cache token spikes, this dashboard gives you the visibility that the Claude Code UI does not provide.
Frequently Asked Questions
1
Where are Claude Code session files stored?
Claude Code writes JSONL session files to ~/.claude/projects/ on your machine. Each project gets its own subdirectory with conversation logs containing token counts, model info, and timestamps.
2
Is my data sent to any server?
No. All processing happens entirely in your browser using JavaScript. Your session files are read locally and never uploaded anywhere. You can verify this in your browser's network tab.
3
How are costs calculated?
Costs are taken directly from the costUSD field in each JSONL entry, which Claude Code records per-turn. These represent API-equivalent costs based on Anthropic's pricing at the time of the session.
4
Can I analyze multiple projects at once?
Yes. You can drop an entire folder or select multiple .jsonl files. The analyzer will process all files and aggregate the data across projects, sessions, and models.
5
What do cache tokens mean?
Cache read tokens are tokens served from Anthropic's prompt cache instead of being reprocessed. Cache creation tokens are tokens written to the cache for future reuse. High cache read ratios indicate efficient context reuse across conversation turns.