Simple setup. Every available feature free.
Claude Code, Codex and Gemini already record their activity. Usage Radar brings the supported counters into one private dashboard. No account, payment details or provider API keys needed for local use.
See the real dashboard with clearly labelled example data. Try the filters and session details before you install anything.
Install the local connector, then open Connect. It reads supported agent logs and serves normalized usage to your browser over loopback.
Check token counts, model costs, cache usage and coverage gaps. Connection status and pricing sources help you judge what each number means.
Install once
On macOS and Linux, the installer below downloads the standalone build, checks its published SHA-256 and sets up the local connector. No Node or npm installation is needed. Windows downloads and step-by-step help are on the connection page.
curl -fsSL https://autopayradar.com/install.sh | sh
The current downloads are not notarized or publisher-signed. Review the installation and verification guide before running them.
Included, free
Usage Radar does not charge for these features. Your AI providers still bill you under their own plans.
The connector processes logs locally. The website receives normalized usage counters, not raw prompts, responses or source files. Public model-price updates may use the network; your local agent logs are not uploaded for those updates.
Costs combine observed token counts with model rates, unless a source records its own cost. Cache reads and writes have separate treatment. Your provider invoice remains the authority on what you were charged.
Coverage depends on the logs available on your machine. Cloud-hosted Codex work leaves no local token log; it is shown as unobservable. Missing usage or an unknown model rate is not presented as a known zero.