PuppyIP Resource Center
AI Tools & Networking 6 minutes Published 2026-11-14

Cloudflare MCP counts dropped? Check response_status

If Cloudflare MCP call counts look lower, check response_status first: the default success filter counts successful calls only, while all includes every status.

Cloudflare One MCP analytics response_status Errors and cached calls

Service eligibility and regional restrictions

PuppyIP serves only compliant overseas businesses and their authorized personnel. Proxy services are not available in mainland China. The service may only be used for lawful business activities outside mainland China. Use of this service within mainland China is prohibited.

Hosting a proxy IP or server overseas does not change these restrictions. The service must not be provided to end users in mainland China through relaying, forwarding, sharing or resale. Before use, read the Terms of Service.

Key Takeaways

  • response_status defaults to success, so new queries omit errors and cached calls; set all to use the all-status counting scope described before the update.
  • The accepted values are success, error, cached, and all. The account, portal, and server tool-call timeseries endpoints use the same parameter, and invalid or empty values return 400.
  • Match daily or monthly granularity, aggregate, time zone, and window when comparing reports. Daily days defaults to 7 and accepts 1 through 179; monthly queries ignore days.
  • The response echoes the selected status and gives a window in epoch milliseconds, including start and excluding end. The example below is hypothetical, not measured traffic.

The success default can make old and new totals look different

Cloudflare’s October 28, 2026 update added response_status to the account, portal, and server MCP tool-call timeseries endpoints. The default is success, so a new query that omits the parameter counts successful calls only.

The update says earlier counts also included failed and cached calls. Comparing an earlier all-status total with a new default result can therefore make the number look lower; that difference alone does not show whether actual call activity fell.

To reproduce the earlier all-status scope, pass response_status=all explicitly. Choose error or cached to count either status separately. The documented values are success, error, cached, and all.

Check the status filter and endpoint scope

The account endpoint returns daily or monthly MCP tool-call counts across the account. For a narrower view, use the server or portal timeseries endpoint. The server path is /accounts/{account_id}/access/ai-controls/mcp/analytics/servers/{server_id}/tool-calls/timeseries; the portal path is /accounts/{account_id}/access/ai-controls/mcp/analytics/portals/{portal_id}/tool-calls/timeseries. All three scopes use the same query parameters.

The response includes the selected response_status. An invalid value, including an empty value, returns HTTP 400 instead of silently reverting to success. If you receive a 400, check the parameter name and use one of the four documented values. When comparing counts, make sure both requests use the same account, portal, or server scope.

Use an explicit all-status GET path

This unexecuted relative GET path illustrates the parameters; account_id is a placeholder.

GET /accounts/{account_id}/access/ai-controls/mcp/analytics/tool-calls/timeseries?granularity=daily&aggregate=false&tz=utc&days=30&response_status=all

The example requests daily, non-aggregated counts in UTC for the trailing 30 days and explicitly includes every status. For one server or portal, use its scoped path with the same parameters. These two sources do not specify actual account permissions, plan fees, or country-by-country availability, so access for every account cannot be assumed.

Match granularity, time zone, and query window

granularity accepts daily or monthly and defaults to daily. aggregate defaults to false; set it to true to request aggregated counts. tz defaults to utc and also accepts Z or a fixed +HH:MM or -HH:MM offset, from -12:00 through +14:00. A fixed offset controls local-day bucketing and does not follow daylight saving time changes within the window.

For daily results, days sets the trailing query window from 1 through 179 and defaults to 7. It is ignored when granularity=monthly. The response echoes granularity, aggregate, tz, and response_status, and includes start, end, series, and total. start and end are epoch milliseconds; the window includes start and excludes end. For daily results, series.day uses YYYY-MM-DD in the requested time zone.

Use hypothetical counts to understand the difference

Suppose a purely illustrative window has 100 calls: 70 success, 20 error, and 10 cached. Omitting response_status or selecting success returns 70; all returns 100; error and cached return 20 and 10 respectively. These figures only demonstrate the filter and are not measured Cloudflare traffic or account data.

For a comparison, hold endpoint scope, granularity, aggregate, tz, and date window constant, then check the response’s response_status, start, and end. Use all when comparing with earlier totals that included every status. Once the query settings match, any remaining difference is a lead for further investigation into call activity.

Sources

Frequently Asked Questions

Does days tell me how long Cloudflare retains analytics data?

No. days is the trailing-window parameter for daily queries; it accepts 1 through 179 and defaults to 7. Monthly queries ignore it. The two official sources reviewed here do not specify a retention period, so the query window is not a retention guarantee.

Does the cached-call count mean I saved the same number of billable calls?

That cannot be concluded from this endpoint. It returns call counts by response status; the reviewed materials do not explain fee treatment or calculate savings. Treat the cached count as analytics data, not as billing evidence or a savings estimate.

Does this update change Skills or Events authorization?

The announcement describes Cloudflare One analytics endpoints and does not announce a Skills or Events authorization change. For those client capabilities and permissions, see the separate Cloudflare API MCP guide: https://puppyip.com/en/resources/cloudflare-api-mcp-skills-client-support-guide.