Local token & cost dashboard for your AI coding tools
Tokdash reads your CLI logs on your own machine and shows exactly how many tokens you burn — and what they cost — across Codex, Claude Code, OpenCode, Gemini CLI and more. No account, no upload, no telemetry.
100% local
Reads logs already on your disk. Nothing is ever uploaded.
Zero telemetry
No accounts, no tracking, no phone-home. Works fully offline.
Open source
MIT licensed and auditable. Install in one command with pipx.
Dashboard installs as a PWA
Pin the dashboard like a native app, light or dark, EN / 中.
One dashboard for every agent
Tokdash unifies usage across nine AI coding tools — switch agents without losing the bill.
Codex
Claude Code
OpenCode
Gemini CLI
OpenClaw
Kimi CLI
Pi
Copilot CLI
HermesEverything you spend, made visible
From raw token counts to real dollars, drilled down to a single session.
Exact token accounting
Input, output, and cache tokens broken out per tool, model, and day — no estimates, the real counts from your logs.
Cost in real money
A built-in pricing database turns tokens into dollars across providers and models, so you see the bill, not just the volume.
Contribution heatmap
A GitHub-style calendar of your activity — with a 3D isometric view and an energy estimate to recolor days by kWh.
Session explorer
Drill into any session, grouped by project and across tools, to find exactly what ran up the cost.
Statusline integration
Surface live token and cost indicators right in your shell prompt while you work — no need to switch windows.
10 themes, light & dark
Classic, Midnight, Terminal, Arcade and more — with system-preference detection, custom date ranges, and English / 中文.
Managed background service
tokdash setup configures a reversible, no-sudo systemd/launchd service in one step — doctor, update, and uninstall keep it healthy and fully removable.
Subscription quota tracking
Monitor remaining Codex, Claude Code, and Antigravity quotas. Track your consumption history, view usage graphs, and set custom update intervals in a single view.
See where every dollar goes
Tokdash breaks spend down to the session, model, and project — so you can find waste and cut it, not just watch the number climb.
-
Cost by model, agent & project
Rank exactly what's running up the bill across every tool.
-
Spot low cache-hit sessions
A low hit rate means you're re-paying full price for context you could reuse.
-
Flag output-heavy turns
Output tokens cost ~5–10× input — see which sessions lean on them.
-
Compare models for cheaper swaps
Find where a lighter model would do the job for a fraction of the cost.
Up and running in 30 seconds
Python package, one command to install, one to set up. Your data never leaves your laptop.
-
1
Install
pipx install tokdash -
2
Setup
Run
tokdash setup. It configures a reversible, no-sudo background service and finds your CLI logs automatically. -
3
Done
The wizard prints your dashboard URL — default
http://127.0.0.1:55423.
Already set up? Upgrade any time with tokdash update.
See it before you install
The live demo runs the real dashboard on synthetic data — explore every tab, theme, and chart right in your browser.