Quickstart

Pick the section that matches your scenario. All three install the same server — they differ only in how you collect transcripts and where you run the dashboard.

Solo — 2 minutes

npm install -g @wigtn/token-server
wigtoken init   # creates ./data/stats.db
wigtoken start  # listens on :10103 by default

Open http://localhost:10103/setup and run the wizard. Paste the bootstrap admin token printed in the server log.

The server defaults to watching ~/.claude/projects/ on the same machine, so any Claude Code session you run will start showing up immediately.

Team — 10 minutes

One person (the operator) runs the server somewhere always-on (laptop docked, mac mini, small VPS):

docker run -d --name wigtoken \
  -p 10103:10103 \
  -v wigtn-data:/data \
  -e MODE=team \
  -e ALLOWED_ORIGINS="https://your-team-site.com" \
  ghcr.io/wigtn/wigtoken:latest

Team members install one of:

Claude Code hooks — paste a PostToolUse block into ~/.claude/settings.json. Setup time: 30 seconds. See hooks.
Agent CLI — npm install -g @wigtoken-temp/agent, then wigtoken-agent run --server https://your-server --token wts_…. More resilient (offline queue + retries). See agent.

The operator gives each member an ingest-scope token from /admin/tokens.

Org — 30 minutes

Same as team, plus:

Run the server behind nginx/cloudflared with TLS.
(Optional) Wire Prometheus to /metrics and import the Grafana dashboard — only if you already run that stack. wigtoken’s own dashboard at / is the primary visualisation surface.
Issue tokens via the admin API (POST /api/admin/tokens) from your provisioning scripts so the dashboard isn’t a manual step.

See deployment → kubernetes for the Helm chart.

Verify it works

curl https://your-wigtoken/api/usage/totals | jq .

You should see messages > 0 after your first Claude Code turn. If not, check the troubleshooting section in deploy.md.

Introduction Deployment