Measure AI coding spend, tokens, and efficiency across every session.
Local-first AI coding performance intelligence for developers and teams.
tokeIT brings every AI coding session into one dashboard so developers and teams can see what they spent, what changed, and what to improve next.
tokeIT brings Claude Code, Codex CLI, Cursor, Gemini, Copilot, and others into one performance layer so developers stop jumping between fragmented logs and token counters.
See every provider through the same session model, cost model, and output model.
Compare Sonnet, GPT, Gemini, and custom models without leaving the same dashboard.
Keep using your preferred tools. tokeIT sits underneath them rather than replacing them.
Show the same week filtered across three providers with visible cost, tokens, commits, and a model breakdown in one frame.
tokeIT should feel frictionless: one install, silent capture in the background, and a desktop view that turns raw sessions into useful performance feedback.
tokeIT.app + background agentInstall the desktop app and the background agent starts watching the AI tool logs already written to disk. No API keys, no timers, no manual start buttons.
Tokens · cost · time · commits · scoreWhen a session ends, tokeIT captures the token breakdown, total cost, duration, git output, and an efficiency score that weighs all of it together.
Stats · trends · insights · coachingThe desktop app shows your trends, highlights waste, and gives coaching on which habits are costing you money and what your best sessions have in common.
Keep the title outside the motion. The animation itself should feel like one full product view opening layer by layer: capture, context, outcome metrics, and coaching.
Show a real capture-first view with provider, model, token buckets, session timing, and project path visible in one clean screen.
This layer should show the raw intake the product understands automatically across tools.
Show a session expanded with git activity, files changed, duration, and model context so the product feels operational, not abstract.
Show how a session becomes meaningful once project context and commit signals are attached.
Best version: one screen with efficiency score, cost per commit, tracked commits, and a compact trend comparison against baseline.
Surface cost per commit, efficiency score, retries, and tracked output in one operational frame.
Show the insights feed with one expanded recommendation, estimated savings, and enough surrounding UI that it feels like a finished product page.
The last layer should feel like calm, high-signal coaching rather than another dashboard widget.
The site should make it clear that the product stays sharp for individual developers while still expanding into team visibility later.
Show the leaderboard, a team trend chart, and a project breakdown in one composed admin screen. That will make the team story feel real immediately.
Leaderboards, per-project tracking, and coaching opportunities belong in one clear admin view. The product should feel like a serious engineering instrument, not gamified surveillance.
The privacy story should feel calm and confident: local storage first, sync only when enabled, private coaching kept private, and shared team analytics scoped to what managers actually need.
The product is built so the sensitive layer remains local unless the user intentionally connects a sync account.
Members can compare shared metrics without exposing session content, private coaching text, or raw prompts.
The shared layer is there to support teams and projects, not to mirror every interaction a developer has with AI.