Who it is forA team lead or finance-minded administrator who needs to see how shared AI capacity is being consumed.
The problemWithout per-key numbers, the team cannot tell which employees use the pool or how much subscription quota remains.
The outcomeA per-key view of input and output tokens and a current reading of remaining subscription capacity and reset timing.
What you need first
- Docker Engine and Docker Compose installed
- openssl available to generate secrets
- AI subscription or API accounts that you are authorized to manage
- Administrator credentials for a running Subpool console (set in .env)
Steps
- Open the Accounts view in the Subpool console and check each account's remaining subscription capacity and reset availability.
- Review the aggregate input and output token telemetry shown for each employee API key.
- Run
curl http://localhost:8080/metricsto read the Prometheus metrics endpoint on the employee API listener. (inferred, check the README) - Note that token figures use provider-native units, so compare them only within the same provider.
Prompt for your agent
Using the gesta-run/subpool repository, query the metrics endpoint at <Subpool API URL>/metrics and list the request counters and token-related series you find. Expected output: a short table of metric names and current values. Verify that each metric name appears in the response body before reporting it.
Caveat. Telemetry is aggregate and in provider-native units, so it does not show per-request content and should not be treated as a cross-provider cost figure.
Use this in your AI agent
Ask your agent
Using the Deduped MCP, get the setup steps for gesta-run/subpool and "Track token use and quota per employee key"
Add Deduped to Claude Code once
claude mcp add --transport http deduped https://dedupedai.com/mcp
Every Monday: the 5 hottest repos and exactly how to use them. Free, one email a week, unsubscribe anytime.
Get them freeShare: X LinkedIn Hacker News Email Copy link
Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer