Universal AI coding cost analyzer — scan Claude Code, Cline, Roo Code, Cursor, Codex CLI, and more for token usage and costs. No API key required. Try it now at aicost.pages.dev.
If you use multiple AI coding tools (Claude Code + Cursor + others), ccusage only covers Claude Code. aicost scans all your AI coding sources in one pass and produces a unified cost report.
| aicost | ccusage | |
|---|---|---|
| Claude Code | Yes | Yes |
| Cline | Yes | No |
| Roo Code | Yes | No |
| Cursor | Yes | No |
| GitHub Copilot | Yes | No |
| Codex CLI | Yes | No |
| Cache-aware pricing | Yes | Partial |
| HTML dashboard | Yes | No |
| CSV export | Yes | No |
| Cost alerting | Yes | No |
| Budget tracking | Yes | No |
| Monthly comparison | Yes | No |
| JSONL export | Yes | No |
| Multi-source unified report | Yes | No |
| Web analyzer | Yes | No |
pip install aicost
# Analyze all detected sources
aicost scan
# Analyze Claude Code sessions
aicost scan --source claude
# Analyze Cline tasks (auto-detected)
aicost scan --source cline
# Analyze Roo Code tasks (auto-detected)
aicost scan --source roo
# Analyze Cursor (export data from Cursor API first)
aicost scan --source cursor --cursor-export ./cursor-export.json
# Analyze GitHub Copilot (export usage from GitHub settings)
aicost scan --source copilot --copilot-export ./copilot-usage.json
# Generate HTML report
aicost scan --html report.html
# Export as JSONL (pipe to other tools)
aicost scan --jsonl
# Analyze Codex CLI sessions (auto-detected at ~/.codex/state)
aicost scan --source codex
# Export as CSV
aicost scan --csv report.csv
# Alert if cost exceeds $50
aicost scan --alert 50
# Track monthly budget with burn rate projection
aicost scan --budget 100
# Compare this month vs last month
aicost scan --compare
# Combine budget tracking + comparison
aicost scan --budget 100 --compare
# View pricing reference
aicost pricing
aicost pricing claude-sonnet-4-20250514- No API key needed — reads local files only, no telemetry
- Multi-source support — Claude Code JSONL, Cline, Roo Code, Cursor API exports, GitHub Copilot, Codex CLI SQLite
- Web analyzer — instant cost analysis at aicost.pages.dev, no installation needed
- Cache-aware pricing — accounts for cache read/write costs that most tools miss
- Budget tracking — set a monthly budget, see burn rate and month-end projection
- Monthly comparison — this month vs last month with per-project/model deltas
- Terminal report — instant cost breakdown with rich formatting
- HTML dashboard — optional visual report with Chart.js graphs
- Cost alerting — terminal + JSON alerts when spending exceeds threshold
- Cost trends — see daily spending over time
- Built-in pricing database — Anthropic, OpenAI, Google, Qwen, DeepSeek, Mistral
| Source | Format | Notes |
|---|---|---|
| Claude Code | JSONL | Auto-detected at ~/.claude/projects/ |
| Cline | JSON | Auto-detected at ~/.cline/state/taskHistory.json |
| Roo Code | JSON | Auto-detected at VS Code extension globalStorage |
| Cursor | JSON/JSONL export | Export from Cursor API or dashboard |
| GitHub Copilot | JSON export | Export from GitHub Copilot settings |
| Codex CLI | SQLite | Auto-detected at ~/.codex/state |
Cursor stores token usage data via their web API (not locally). To analyze Cursor usage:
- Export usage data from Cursor's dashboard or API
- Save as JSON/JSONL — any file in the Cursor API response format
- Run:
aicost scan --cursor-export ./cursor-usage.json
The parser accepts:
- Single event JSON:
{"timestamp": ..., "model": ..., "tokenUsage": {...}} - API response JSON:
{"usageEventsDisplay": [...]} - JSONL: one event per line
GitHub Copilot usage data can be exported from your Copilot settings:
- Export usage data from GitHub Copilot settings
- Save as JSON — the exported file contains usage events with token counts
- Run:
aicost scan --source copilot --copilot-export ./copilot-usage.json
The parser accepts:
- Array of events:
[{"model": "...", "prompt_tokens": N, "completion_tokens": N}] - Wrapped response:
{"data": [...]} - JSONL: one event per line
Cline stores task history as a JSON array in ~/.cline/state/taskHistory.json (CLI) or the VS Code extension globalStorage directory. Each task record includes token counts, cache usage, and cost:
- Auto-detected — run
aicost scan --source clineoraicost scan(all sources) - Custom path —
aicost scan --source cline --path /custom/taskHistory.json
The parser reads: tokensIn, tokensOut, cacheWrites, cacheReads, totalCost, modelId, and project path.
Roo Code (forked from Cline) stores task history in the same format at its VS Code extension globalStorage directory:
- Auto-detected — run
aicost scan --source roooraicost scan(all sources) - Custom path —
aicost scan --source roo --path /custom/path
Both Cline and Roo Code use identical HistoryItem schema, so one parser covers both tools.
Codex CLI stores thread metadata in a SQLite database at ~/.codex/state:
- Auto-detected — run
aicost scan --source codexoraicost scan(all sources) - Custom path —
aicost scan --source codex --path /custom/state
Note: Codex stores total tokens_used without input/output split. We estimate 70% input / 30% output (typical for coding agents) and calculate cost from model pricing.
No installation? No problem. Visit aicost.pages.dev to:
- Upload your AI tool data files (JSON/JSONL)
- Paste raw data directly
- Instant analysis with cost breakdown by project and model
- Export CSV for spreadsheets
- 100% client-side — your data never leaves your browser
AI Coding Cost Report
=====================
Period: 2026-04-01 to 2026-04-16
Total Cost: $12.34
Total Tokens: 1,234,567
Input: 987,654
Output: 246,913
By Project:
pr-explainer-action $4.56 (37.0%)
ogpixel $3.21 (26.0%)
By Model:
qwen3.6-plus $8.90 (72.1%)
sonnet $2.34 (19.0%)
Daily Trend:
2026-04-14 $3.45 ████████████████
2026-04-15 $5.67 ██████████████████████████
2026-04-16 $3.22 ███████████████
Pricing database includes: Anthropic Claude, OpenAI GPT/o-series, Google Gemini, Qwen, DeepSeek, Mistral.
See all pricing reference: aicost pricing
Look up a specific model: aicost pricing claude-sonnet-4-20250514
Set a spending threshold to catch cost overruns:
aicost scan --alert 50If your total cost exceeds $50, aicost prints a red alert and outputs a structured JSON alert to stderr for piping to monitoring tools.
Export individual usage entries as JSONL for use in other tools (spreadsheets, monitoring, custom dashboards):
aicost scan --jsonl > usage.jsonlEach line is a JSON object with timestamp, model, project, token counts, and cost.
Export your usage data as a CSV file for analysis in spreadsheets:
aicost scan --csv report.csvThe CSV includes: timestamp, model, project, source, token counts, and cost. Plus summary sections for totals, by-project, and by-model breakdowns.
Set a monthly budget to track spending, see your daily burn rate, and get a projected month-end cost:
aicost scan --budget 100Output includes:
- Remaining budget — how much you have left
- Daily burn rate — average daily spend
- Projected month-end — estimated total if current pace continues
- Status indicator — green (under), yellow (warning at >75%), red (over budget)
Compare this month's AI coding costs against last month:
aicost scan --compareShows:
- Cost delta — absolute and percentage change
- Token delta — usage volume change
- Per-project changes — which projects cost more/less
- Per-model changes — spending shift across models