Tokenscope
Exposes three MCP tools that parse your local Claude Code session logs to show token cost attribution. The analyze_claude_cost tool breaks down spend by output versus context re-sent every turn (typically 60%+ of bills), traces context growth across the session, and flags which models or tools are expensive. get_cost_benchmark compares your session against an offline reference dataset. tokenscope_share_summary generates privacy-safe markdown and SVG reports with no file paths or prompt content. Reads from ~/.claude/projects/**/*.jsonl, strictly local and read-only. Useful when an AI bill surprises you and you want the agent itself to explain where the money went, or when you're optimizing long-running sessions and need per-turn context metrics without leaving the chat.
- GitHub stars
- 4
- tools
- 3
- transport
- STDIO
- last updated
- Jul 25, 2026
Tools
3 toolsPublic tool metadata for what this MCP can expose to an agent.
analyze_claude_costAnalyze a Claude Code session for cost and context attribution.
Analyze a Claude Code session for cost and context attribution.
No parameter schema in public metadata yet.
get_cost_benchmarkCompare a session's cache efficiency against the offline reference set of real sessions.
Compare a session's cache efficiency against the offline reference set of real sessions.
No parameter schema in public metadata yet.
tokenscope_share_summaryProduce a privacy-safe shareable summary of a session's cost breakdown.
Produce a privacy-safe shareable summary of a session's cost breakdown.
No parameter schema in public metadata yet.
Sponsors
Agent CaffeineSponsoredClose your lid, come back to finished work
SnugSponsoredThe clipboard manager for the AI eraPromote your product hereReach 125k+ visitors monthly
WaxSponsoredBlock anything with supervision, without erasing your iPhonePromote your product hereReach 125k+ visitors monthly
knobsSponsoredSee your site on any phone, any day, in any languageReadme
tokenscope ⏣
See what your AI-coding session actually cost — and what's eating your context. A local, read-only CLI that parses your Claude Code session logs and shows where the money goes: model output vs. context being re-sent every turn (the hidden 60%+ of most bills).
$ npx @wartzar-bee/tokenscope
tokenscope ⏣ latest session
──────────────────────────────────────────────────────
Total cost $868.84 over 967 model turns
Where the money went
output (model writing) ████░░░░░░░░░░░░░░░░░░░░ 16% $137.24
cache read (re-sent ctx) ████████████████░░░░░░░░ 66% $577.59
cache write (new ctx) ████░░░░░░░░░░░░░░░░░░░░ 18% $153.67
Context size per turn (peak 822k · avg 404k · now 822k tokens)
▁▁▁▁▁▂▂▂▂▂▂▂▃▃▃▃▃▃▃▄▄▄▄▄▅▅▅▅▅▆▆▆▆▇▇▇▇▇▇█
Insights
• Re-sent (cached) context cost $577.59 (66% of spend) — context re-read every turn.
• Peak context ~822k tokens — /compact or a fresh session would cut per-turn cost.
• Only 16% of spend is the model's actual output.
(A real session, default Opus pricing. Your numbers will differ — prices are overridable.)
Details
- Repository
- github.com/wartzar-bee/tokenscope
- Website
- github.com/wartzar-bee/tokenscope#readme
- Package
@wartzar-bee/tokenscope-mcp- Transport
- STDIO
- Registry
- MCP registry, active