Analytics

Usage and billing

Review customer-scoped workspace runtime, managed Archibot AI activity, platform resource usage, reconciliation posture, and billing readiness across the Analytics tabs, and set spending caps.

Customer adminsCustomer membersPlatform operators

Last updated June 18, 2026

Analytics is where you review how a customer account is using Archibot: workspace runtime, managed Archibot AI, platform resources such as CI and Shared Drive, reconciliation posture, and billing readiness. It lives at Analytics in the Console navigation and is the same surface support uses for billing handoff.

This guide describes the page as it renders today: a fixed 30-day window, a shared filter bar, role-gated metric cards, seven tabs, and a Spending caps tab that writes monthly caps back to the usage ledger.

Who can view usage

What you see depends on your role.

If your account has no tenant membership yet, Analytics tells you so and waits until onboarding creates a tenant before it scopes usage. Account access and role mapping are covered in Access roles.

The window and filter bar

Analytics always reports a fixed 30-day window. There is no period picker; the snapshot is the trailing 30 days, and the active scope and window are summarized in the Window snapshot panel on the right.

At the top of the page a shared bar drives every tab:

• Search filters users, services, workspaces, and usage rows.
• The service filter (the all control) narrows to a single billable service, for example workspace runtime or managed AI.
• The customer selector (admins and operators only) chooses All visible customers or one customer account. Members do not see this selector because they are already scoped to themselves.
• Refresh reloads usage and spending caps.

The customer you pick in this bar is the scope for the Members, Workspaces, and Spending caps tabs. Confirm it before reading or changing anything.

Key metrics

Below the filter bar, the Key metrics cards summarize the window. The cards shown depend on your role:

You can collapse this block with Hide metrics and reopen it with Show metrics.

The seven tabs

Analytics organizes detail into tabs. Two of them (Members and Spending caps) are hidden for members.

The third tab is labeled Costs when you can see billable amounts and Usage for members who cannot.

Overview and Trends

Overview opens by default. It shows the billable-estimate-over-time chart (or usage-over-time for members) and a current-window summary describing the day count, scope, and coverage. Trends breaks the same data into a daily series so you can read momentum and catch a spike before opening individual line items.

Costs

The Costs tab (shown above) lists each charge line with its usage and, for admins and operators, a billable estimate and included or over-allowance state. Use the table search to narrow rows.

Two callouts may appear here:

• A current bill estimate banner restates the window total. This is showback guidance for the selected window; Stripe meter exports and prepaid debits remain controlled by backend jobs.
• For billing operators, a gross margin callout compares known-cost margin to the target.

An Accuracy Disclaimers card explains the cost-math rules that keep estimates honest, for example that failed managed-AI calls without usage counters are not debited, and that managed AI is included in billable estimates only when reconciliation is complete for the visible scope.

A Resource Showback card lists metadata-only usage for CI, review, QA, persistent environments, and Shared Drive, with actor attribution and resource allowances. If any rows are still unpriced, an amber note names the resource kinds that need pricing.

Members

The Members tab groups workspace runtime, managed AI, and resource events by member for the selected window and customer. Use it to see which members drive activity. It is hidden for member-role accounts.

Workspaces

The Workspaces tab shows a runtime-over-time chart and a Runtime Consensus panel comparing the Kubernetes ledger hours, the Coder build-history comparison, and the drift between them. Below that, the Workspace Breakdown table lists each workspace with its owner, target, runtime, billable estimate (for admins and operators), and last-observed time.

Runtime is shown from the billing-grade Kubernetes ledger when available and falls back to Coder build history otherwise; the table labels the source.

Spending caps

The Spending caps tab is a write surface for customer admins and operators. Select a single customer account in the filter bar first; with All visible customers selected the tab asks you to choose one account.

Each row is one service cap. For every row you can set:

To change a cap, edit its fields and choose Save on that row. Each row saves independently, and the save button stays disabled until you change something. Console stores these policies alongside the usage ledger so billing and launch gates read one source.

A separate Data gaps card lists services that still need attribution or pricing before a cap can become hard enforcement. If a service shows a missing-data badge, set the cap action to Alert only until the underlying data is complete; Block new work depends on accurate attribution.

Data Safety

The Data Safety tab documents what Analytics tracks and what it never touches.

• Where We Track It lists the storage layers for estimates, rollups, and billed-cost allocation, each with a posture badge.
• Explicitly Excluded confirms that raw prompts, model responses, full transcripts, raw provider keys, and private policy assets stay out of analytics and invoice-support exports.
• Event Contract shows the minimum metadata fields kept for customer-safe analytics.
• Live Proxy Usage shows estimated usage from the Archibot proxy ledger, which is intentionally separate from reconciled provider billing.

Treat usage data as account evidence. It should never include prompts, responses, transcripts, raw provider payloads, API keys, cookies, or workspace secret values.

Reconciliation posture

Provider reconciliation can lag provider billing systems by about a day, and Admin API windows settle after the fact. The Window snapshot and provider posture panels on the right show the current reconciliation state.

If a recent number looks low or missing, check the reconciliation posture before escalating. Managed AI is included in billable estimates only once reconciliation is complete for the visible customer scope, so an unreconciled window can read low by design.

Why provider IDs are hidden

Customer-scoped Analytics intentionally hides provider project IDs and secret-bearing provider metadata. That is expected. The page is built to give useful usage and billing status without exposing operational identifiers. Operators who need mapped-project detail see it only in operator scope.

Billing readiness and workspace creation

Workspace creation depends on the account’s billing state. Verified paid access or an approved trial can allow create when other readiness gates are clear. Prepaid accounts also need funded workspace-time and managed-AI balances before create proceeds. Catalog funding and the readiness gates are covered in Catalog and readiness.

When you ask billing a question

For a formal billing question, include the customer account, the 30-day window, the visible totals, and the reconciliation posture. Do not include payment secrets, invoice payloads, provider account secrets, or webhook details. The Support handoff guide describes what to gather, and Troubleshooting covers usage that looks wrong before it reaches support.

Related guides

• Catalog and readiness — funding and the gates that allow workspace creation.
• Access roles — which roles see customer-wide totals and the Spending caps write surface.
• Tenant Admin for operators — operator-scoped account and tenant management.
• Support handoff — what to collect before opening a billing question.
• Troubleshooting — diagnosing usage or readiness that looks wrong.