Server analytics tools
Traffic, clients, tools, health, countries and error logs for one server.
Six read-only tools answer questions about one server's traffic. They read the same data as the Analytics tab and each takes the server's serverId (from list_servers). The first five take a time window of 24 hours, 7, 30 or 90 days; get_server_analytics_error_runtime_logs takes the time of the errored event instead.
| Tool | Returns |
|---|---|
get_server_analytics_summary | Requests, sessions, tool calls, errors and latency for the window, with the change against the previous window of the same length |
get_server_analytics_client_breakdown | Which MCP clients connect, with sessions and requests per client over time |
get_server_analytics_tool_breakdown | Calls per tool, prompt and resource over time |
get_server_analytics_health | Per-tool and per-client error rates, tool-result errors and p50 / p95 / p99 latency |
get_server_analytics_requests_by_country | Requests and sessions by country |
get_server_analytics_error_runtime_logs | Runtime logs around an errored request, from the deployment that served it |
A session is one conversation or connection. By default Manufact's own clients (the dashboard, the hosted inspector and the publish checks) are left out, so the numbers show real user traffic.
Retention by plan
The window is limited by your plan's analytics retention: a longer window is shortened to what your plan keeps, as in the dashboard. See Billing & plans.
Example prompts
- "How much traffic did my-server get this week, compared to last week?"
- "Which clients call my-server most, and which tools do they use?"
- "Which tools have the highest error rate over the last 30 days?"
- "Show the logs around the latest errors of my-server."
Related
list_servers: find theserverIdget_deployment_runtime_logs: full runtime logs of a deployment