Usage and reporting limits

Open Manage → Usage & budget with billing:view. The page reports private ingestion, total accepted telemetry, public-agent consumption, estimated cost, reporting capacity and budget events. These are metered estimates, not invoices or payment transactions.

The page has three tabs. Billing contains the current estimate, budget outlook and account controls. Usage contains ingestion history and public usage details. Metric queries contains the daily database scan allowances and query activity. The selected tab is preserved in the page URL.

Read the period and history

Usage follows UTC calendar months. Daily history covers recorded days in the current month. Earlier usage without daily attribution remains identified as undated; missing dates are not invented zero-usage days.

Reporting hours are based on leases, with up to two minutes reserved after the last successful request. Overlapping renewals are not counted twice. Telemetry counts accepted normalized measurement data, rather than network packet sizes. Uploading old samples does not reconstruct historical reporting hours.

Public agents have a separate accepted-byte allowance and dedicated per-GB rate that includes hosted compute. Public bytes are not also charged as private data. No private reporting hours or public check-hours are billed. Public terms must be provisioned by the provider before enforced public access is available.

Budget outlook shows a month-end scenario at the month-to-date average pace. It is not a guaranteed bill or a forecast of future workload changes. An early month may not have enough history to show one.

Dedicated and shared usage

The breakdown separates dedicated ingestion, your allocated shared telemetry, public HTTP response traffic and the monthly base. It also shows the full measurement content represented by your shared observations and the bytes saved by allocation. Savings compare content bytes, not physical disk usage.

Eligible shared cluster probes and coalesced HTTP, DNS, TCP and direct-echo reports split one physical measurement equally across distinct tenant accounts. Extra checks in one account do not multiply its share. Per-check metadata and destination-specific measurements remain dedicated. Allocated shared telemetry starts at $0.50 per decimal GB for Team and Business, configurable by the provider; this is a separate rate from public HTTP response-body traffic. Public response bytes from a shared report are also divided across its beneficiaries.

Shares remain provisional for 24 hours after the first accepted report. A new subscriber within that window reduces earlier accounts’ shares through credits booked to the original UTC day and month. New subscribers after settlement are refused; existing reference retries do not add usage. Private and public allowances remain separate, and each covers dedicated ingestion before shared ingestion. Monthly price snapshots stay fixed. Replication, indexes and backup copies do not multiply billable ingestion.

For example, one eligible 1,000-byte measurement shared by ten accounts allocates 100 bytes to each account, plus its own dedicated metadata. Adding another check in one of those accounts does not create another account share. Sharing requires an actual common observation; matching destination or router IP alone is insufficient.

Month-end shared usage can change during the settlement window. Invoice close must wait until that window finishes. Operators can reconcile the usage ledger; these estimates do not collect payment.

Metric query allowance

This section separates daily scan allowances for interactive and assistant queries. It shows remaining, debited and reserved bytes, outstanding requests and the reset time in your local timezone. Reservations can conservatively account for interrupted work; these safety debits are not billing charges.

If an allowance is exhausted, wait for outstanding requests to settle or the daily reset. Narrow subsequent time windows and filters to reduce scanning. The section refreshes every 30 seconds and has a retry action if unavailable.

Query activity groups repeated requests by a fingerprint, origin and time-window duration. A moving one-hour window remains one group; changing filters or the window duration produces another group. A run is one API lookup, which may execute multiple SQL statements. Exact duplicate panels in a single dashboard request share execution while retaining their separate results.

Choose today or seven UTC calendar days and filter by interactive or assistant origin. Groups are paginated in descending safety-debit order. Expand a group for SQL execution counts, request time, last run, outstanding/abandoned counts and a one-run-per-minute daily scenario. Scan averages divide observed bytes by settled requests, including failures and abandoned work; interrupted reads can be incomplete. Dollar cost is Not priced: safety debits are not charges. Only seven days of detailed history are retained, and earlier entries without fingerprints cannot be split into individual repeat queries. The journal stores route patterns and identifier-like metric names, not query text or filters.

Live chart refresh follows its bucket width, between five and sixty seconds; long historical charts therefore rescan less often. Fixed-time chart windows do not poll. An execution-budget error stops automatic retries until you adjust or explicitly retry the query. A daily-budget error waits for the server’s reset time. Dashboards use the interactive allowance, not the assistant’s.

Change a policy

With billing:manage, use Account guardrails to set the monthly budget, reporting capacity, overage setting, enforcement mode and ephemeral-instance policy. Review the current plan and save the guardrails. Changes are audited. Rate-plan changes through the supported API or AI tool require an unused UTC billing month and do not purchase or provision a commercial subscription.

Observation mode measures usage without enforcing financial or capacity caps. Enforcement controls reporting admission. Lowering capacity below active leases can prevent renewal until capacity becomes available. Ingestion caps refuse new data but do not block private lease renewal. Reporting time has no charge. Refused agents may continue measuring locally, and their bounded spools can eventually fill. A reporting limit is not a promise that local checks stop or that all offline data will be retained.

Threshold events appear in the app. They do not automatically send budget email or webhook notifications. Use budget planning for annual scenarios and cost assumptions, and agent management for reporting controls.