AI Spend

Every agent, one meter.

You run Claude Code, Codex, and Copilot side by side — three tools, three different meters, no single number. This is the cross-tool view: paste what each one gives you (or just type the totals) and see the whole bill in one place. Nothing leaves your browser — the parsing is JavaScript, the numbers are yours.

$0.00total, all tools
0total tokens
0tools metered
—biggest line
① Claude Code — paste session JSONL

From ~/.claude/projects/<project>/*.jsonl. Paste one or many files' lines — this reads the usage fields (input/output/cache tokens) and prices them at the rates below. (This is the same data ccusage reads; here it's one input among several.)

② Any tool — paste a CSV

Two columns any tool exports or you assemble: label,cost (e.g. Codex,12.40) — or label,input_tokens,output_tokens to price by tokens. Header row optional.

③ Manual meters — the flat subscriptions

The ones with no export: type them straight in. label,monthly$ per line.

Claude price ($/M):

🔒 100% in your browser. No upload, no account, no key — paste, meter, done. Prices are yours to set (they change; today's go in the boxes). Claude token costs are computed from real usage fields; CSV/manual are whatever you enter. Estimates, not an invoice.
Want to shrink these numbers? The pm-tokens bundle + the Token Dashboard cut what rides your context; pm-claude-skills mcp-audit finds the MCP servers charging rent.