Usage
A Pi Durable Platform team's model usage: requests, tokens and estimated cost per day, model or agent, where the numbers come from, and exporting them.
The team's Usage page in Pi Durable Platform shows how much the team's agents used their models: requests, input and output tokens, cache reads and writes, and an Estimated cost. Every member can see it, and so can the team's service accounts through the API.
- Period: the date range (days in UTC), up to 366 days; the last 30 days by default.
- Group by: Day, Model or Agent.
- A chart of tokens per day, and a table with the totals.
An agent's Overview shows its own usage over the last 30 days.
Where the numbers come from
- Proxied models: the platform's model gateway relays every request and records the tokens the provider reports.
- Direct-mode models and the platform's default model: the gateway doesn't see these requests, so each agent counts its finished responses and reports them to the platform. Reports for proxied models are ignored, so nothing counts twice.
Usage counts model requests only. Running agents, storage and MCP calls aren't in it.
Estimated cost
The cost uses each model's current prices (US dollars per million input and output tokens), which platform admins set. If a price changes, the whole period is recalculated with the new one. Requests to models without prices are counted but not in the cost; the report says how many.
Service-managed agents
Agents that apps created for their users are named in the report for the team's owners and admins and its service accounts. Other members see them summed in one row.
Through the API
curl "https://agents.example.com/v1/teams/support/usage?groupBy=agent&from=2026-10-01&to=2026-10-31" \
-H "Authorization: Bearer $PDP_SA_KEY"groupBy is day, model or agent, one at a time. Add agentId to see one agent, for example by day. See getTeamUsage.