Usage and logs
Track Ostra requests, tokens, cache hits and spend by model, task, channel and key on the Usage page, export it as CSV, and inspect individual requests on the Logs page.
Usage
Usage (console.testrelic.ai/usage) answers "what did we use and what did it cost?".
Range
This month (default), Last 7 days, Last 30 days or Last month. All dates are UTC.
Summary
| Card | Shows |
|---|---|
| Requests | All requests, and how many Ostra Decision billable units they used. |
| Ostra Pro tokens | Input and output tokens. |
| Spend | Spend in the range. On This month it also shows your monthly spend limit; otherwise a reminder that cached results and errors are free. |
| Via MCP | Requests made over MCP. |
| Cache hits | Ostra Decision answers served free from the 24-hour cache. |
Chart and breakdown
The Daily spend / Daily requests chart (toggle Spend or Requests) groups by Model (default), Task, Channel or API key, showing the top five and "Other".
The Cost by <group> table below lists, for each group, billable units or requests, the split by channel (REST, MCP, Console), the cost, and a total. Playground use appears under API key as "Console (no key)".
Channels:
- REST: calls to
https://ostra.testrelic.ai/api/v1. - MCP: tool calls through the MCP server.
- Console: runs in the Playground.
Export
Export CSV downloads ostra-usage-<from>-to-<to>-by-<group>.csv with one row per day and group:
date,<group>,requests,units,cost_usd,rest,mcp,consoleLogs
Logs (console.testrelic.ai/logs) lists individual requests, newest first, from every channel.
Filters
- Search request ID, key, client or error. Prefixes narrow the search:
tool:,client:,key:,request_id:(for examplerequest_id:req_01JD…). - Status: All statuses, Success, Errors.
- Channel: All channels, REST API, MCP, Console.
- Model and task: All models and tasks, each Ostra Decision task, or Ostra Pro.
Filters are kept in the page URL, so you can share a filtered view. They apply to the newest 2,000 entries.
The table
Time (UTC), status, channel, endpoint or tool, task, key, the decision (or tokens for Ostra Pro), cost ("cached" or "$0" when free) and latency. Show 10, 25 (default), 50 or 100 per page.
Request detail
Click a row for the Request detail drawer:
- Status, endpoint and request ID (copy it when contacting support), and the time.
- Model and task, channel, key ("Console session" for Playground runs), MCP client, the decision or tokens, and what was billed.
- Timing: time in the gateway and in the model.
- Error type and message, for failures.
- The request and response JSON, stored up to 16 KB each.
- Actions: Open in playground (Ostra Decision), Open Ostra Pro playground, and Copy as cURL (REST requests).
If the workspace doesn't store bodies, the drawer says: "Request and response bodies aren't stored for this workspace. Change this in Settings › Data and privacy".
Retention
Logs follow Settings → Data and privacy:
| Setting | Effect |
|---|---|
| 30 days (default) | Logs older than 30 days are deleted. |
| 7 days | Logs older than 7 days are deleted. |
| Zero retention | No request logs are kept. Usage and billing still count every call. Playground history is off. |
| Store request and response bodies | On by default (16 KB each). Off keeps only time, task, status, latency, cost and key. |
Owners and admins can also Delete all request logs there. Usage, billing and the audit log aren't affected. See Settings.
Find a request from your code
Every API response carries an x-request-id header (and errors repeat it as error.request_id). Log it, then paste it into the Logs search to see exactly what Ostra received and returned.
MCP servers
Connect Claude Code, Cursor, VS Code, the Claude API or any MCP client to the Ostra MCP server from the TestRelic Console, choose which tools to expose, and monitor clients and tool calls.
Rate limits
See your workspace's usage level and requests-per-minute limit on the Rate limits page, how levels go up with top-ups, Ostra Decision per-task limits, and per-key overrides.