TokenSpend AI
Arrow keys to move · F for full screen
Usage intelligence for AI coding

See what your AI coding costs.

By project, task and commit, across every assistant you use. It runs on your machine, with no account and no telemetry.

LocalNo accountNo telemetryComing soon to the VS Code Marketplace
SAMPLE DATA
The problem

AI spend is scattered, and it never says which work it paid for.

Teams now use several AI coding assistants at once. The cost is real, but it is hard to see where it goes.

Scattered

Every assistant keeps its own usage logs, in its own format, in its own folder.

Too coarse

Bills arrive by day and by model, not by project, repository or commit.

Disconnected

Nothing ties the money to the task or the change it produced, so it is hard to say what an output cost.

What it does SAMPLE DATA

One local dashboard for every AI coding tool.

It reads the logs already on your disk and turns them into cost you can act on.

Burn vs budgetEstimate

69% of the $600 budget · 55% of the month gone

Cache health

376K of 480K input tokens from cache · Healthy

Cost per commitMeter + estimate

$412 over 145 commits

Reads five toolsSpend by project, task and commitBudgets and forecastsAdvice with monthly savingsExecutive and management reportsCSV and JSON exports
What sets it apart

Honest numbers, tied to the work.

Every number says how sure it is

Each figure is labelled Estimate, GitHub meter or Matched to bill. A model with no published price stays unpriced. It is never counted as $0.

GitHub’s own meter, without admin rights

It reads the per-call charge GitHub records in Copilot CLI logs, so seats paid by an organisation still get real figures. A charge record, not an invoice.

Spend tied to outputs

Cost per repository, commit and task. Start a timer and the AI requests made in that project while it runs are attributed to the task.

Local and metadata-only

No account, no server, no telemetry. It stores model, time, token counts, project and cost. Never prompts, replies or code.

EstimateGitHub meterMatched to bill
Built for developers and IT teams

Nothing to set up. Easy to approve.

For developers

It finds the logs already on your disk. Install it, work as usual, and the cost appears by project, task and commit.

For IT and finance

Runs on the developer’s machine and listens on 127.0.0.1 only. GitHub billing is optional and read-only. Reports export to PDF, CSV and JSON.

Coming soon to the VS Code Marketplacetokenspendai.com
Meet the Creator: Jonathan Davidar (opens davidar.org in a new tab) I solve problems with creative vision and AI solutions architecture.

Verified on Windows so far. There is no team-wide rollup yet. The security hardening has not had an independent review. TokenSpend AI is independent of GitHub, Microsoft, Anthropic, OpenAI, Alibaba, Google and Continue. License terms