What 0token does
Token Flamegraph
See exactly where tokens go — per request, per model, before vs after compression
Live $ Meter
Real-time animated counter showing dollars saved as your agent works
Receipt Cards
Shareable SVG cards — download, copy, post on X
MCP Server
4 tools for any MCP-compatible agent — stats, recent, flamegraph, receipt
Leaderboard
Cross-tool efficiency comparison — $/task at held quality
Zero Config
Works with Claude Code, Cursor, Codex, any OpenAI-compatible client
How it works
Steps
1
Install:
pip install 0token2
Start:
python -m zt.cli start3
Use your agent normally — 0token compresses in background
4
Watch savings on this dashboard
Quick Setup
▸ Claude Code:
set ANTHROPIC_BASE_URL=http://127.0.0.1:8787 && claude
▸ Cursor / OpenAI:
set OPENAI_BASE_URL=http://127.0.0.1:8787/v1
Proof — real numbers
| Workload | Before | After | Savings | Cost |
|---|---|---|---|---|
| 20 mixed requests | 430,000 | 121,500 | 71.7% | $1.175 |
| gpt-4o (10 reqs) | 229,900 | 61,800 | 73.1% | $0.575 |
| claude-sonnet-4 (10 reqs) | 200,100 | 59,700 | 70.2% | $0.600 |