Vaudit runs a local TokenID proxy that records LLM API usage on your machine. Track tokens, latency, timestamps, and cost estimates in a localhost dashboard while prompt and completion content stays local.
Install TokenID, start the local proxy, verify by magic link, then point your AI SDK at the printed localhost URL.
Use Homebrew, apt, dnf, Docker, Windows direct download, or a direct Linux/macOS binary.
Run tokenid start, confirm the magic-link email, and TokenID writes your license to ~/.tokenid/license.json.
Point your AI SDK at the localhost proxy URL. Usage appears in the local dashboard immediately.
The local proxy captures model, token counts, latency, timestamp, and cost estimate for every supported LLM API call.
View live usage, cost trends, model breakdowns, and leaderboards from http://localhost:<port>.
Works with OpenAI, Anthropic, and other providers that can route HTTP traffic through localhost. Streaming and non-streaming calls are captured.
Tool-use events are recorded separately from message events so agent behavior is easier to audit.
Install on macOS, Linux, Windows, or Docker. Debian and RPM packages are GPG-signed with published package keys.
Prompts, completions, token counts, and cost estimates all stay in local SQLite. We collect anonymous usage statistics, tied to the email you verified with, to keep the binary stable and ship fixes. Disable with tokenid --no-telemetry — that also turns off self-upgrade and entitlement offers, since both use the same signal.
Select an operating system or runtime, then use the matching install method. The shared first-run flow is below the tabs.
Recommended for most macOS users. Homebrew manages the binary and keeps updates simple.
brew tap blokid/tap
brew install tokenidUse this when you do not want Homebrew. Choose the architecture that matches your Mac.
PLATFORM=darwin-arm64 # or darwin-amd64
curl -fsSL "https://github.com/blokid/tokenid-api-wrapper/releases/latest/download/tokenid-${PLATFORM}" \
-o /usr/local/bin/tokenid
chmod +x /usr/local/bin/tokenid
tokenid versionUse the signed apt repository on Debian and Ubuntu systems.
curl -fsSL https://deb.token.audit.id/key.gpg | sudo gpg --dearmor -o /etc/apt/keyrings/tokenid.gpg
echo "deb [signed-by=/etc/apt/keyrings/tokenid.gpg] https://deb.token.audit.id/ ./" \
| sudo tee /etc/apt/sources.list.d/tokenid.list
sudo apt update
sudo apt install tokenidUse the signed RPM repository for Fedora, RHEL-compatible, AlmaLinux, and Rocky Linux systems.
sudo rpm --import https://rpm.token.audit.id/key.gpg
cat <<EOF | sudo tee /etc/yum.repos.d/tokenid.repo
[tokenid]
name=TokenID
baseurl=https://rpm.token.audit.id/
enabled=1
gpgcheck=1
gpgkey=https://rpm.token.audit.id/key.gpg
EOF
sudo dnf install tokenidUse this on Linux distributions where you prefer not to configure a package repository.
PLATFORM=linux-amd64 # or linux-arm64
curl -fsSL "https://github.com/blokid/tokenid-api-wrapper/releases/latest/download/tokenid-${PLATFORM}" \
-o /usr/local/bin/tokenid
chmod +x /usr/local/bin/tokenid
tokenid versionOptional verification for the TokenID package signing key. The expected fingerprint is 7B86AC2F28D512FA409B942D6F74126529D250C2.
curl -fsSL https://deb.token.audit.id/key.gpg | gpg --show-keys --with-fingerprintDownload the Windows amd64 executable with PowerShell, then run the version command to confirm it works.
$url = "https://github.com/blokid/tokenid-api-wrapper/releases/latest/download/tokenid-windows-amd64.exe"
Invoke-WebRequest -Uri $url -OutFile $env:USERPROFILE\tokenid.exe
& $env:USERPROFILE\tokenid.exe versionUse the multi-arch Docker image when you want an isolated runtime. Mount ~/.tokenid so the license and local state persist between runs.
docker pull tokenid/tokenid:latest
docker run --rm -it -v ~/.tokenid:/root/.tokenid tokenid/tokenid:latestAfter installation, start TokenID once. It prompts for email, sends a magic link, writes the license to ~/.tokenid/license.json, starts the local proxy, and prints the localhost dashboard URL.
tokenid startLive package repositories: deb.token.audit.id and rpm.token.audit.id.
Nothing about your LLM calls. Prompts, completions, token counts, model names, and cost estimates all stay in local SQLite at ~/.tokenid/data.db. We do collect anonymous usage and crash statistics, tied to the email you verified with, so we can keep the binary stable, ship bug fixes as they happen, and identify enterprise usage patterns for sales outreach. No LLM content of any kind is included. Opt out with tokenid --no-telemetry — this also disables automatic self-upgrade and entitlement offers, since both depend on the same usage signal. See the privacy policy for details.
From the anonymous usage statistics tied to your verified email. If your install shows enterprise-scale activity, we may reach out to discuss a plan that matches your scale. The signal is purely product analytics — no LLM content, ever.
OpenAI, Anthropic, and other providers that use a localhost-routable HTTP endpoint. Streaming and non-streaming responses are both captured.
Yes. The free tier includes local measurement, the localhost dashboard, signed packages, email verification, and a license token that renews automatically while the proxy is online.
After magic-link verification, TokenID writes the local license file to ~/.tokenid/license.json.