Usage

The Usage page reports what your tasks consumed: credits spent on VM runtime, VM uptime, LLM tokens, and agent runs. It reads from a single endpoint and presents the same data two ways — a glanceable Overview and a deep Explore view. Open it under Settings → Usage. Usage attributes everything to tasks. Credits shown here are VM-runtime credits only; tokens are reported for insight but never billed. For your full charged total, including warm-slot snapshot refreshes, see Billing and Billing & Credits.

Overview vs Explore

The page has two tabs. Overview is the default landing surface; Explore is opened explicitly.

Tab	What it shows
Overview	A glanceable dashboard: KPI cards (credits, VM runtime, tokens), a usage-by-time chart, and breakdowns by model and project. Each card drills into Explore pre-pivoted to that dimension.
Explore	The full explorer: pick a metric, group by any dimension, filter, sort and page the task ledger, and export CSV.

Both tabs share the same filtered window, so switching tabs keeps your date range and filters.

Metrics

The KPI cards and charts surface four metrics. Credits is the money metric, runtime is what drives it, and tokens are work done — never billed.

Metric	Field	Meaning
Credits	`credits_used`	VM-runtime cost, derived only from runtime intervals. The primary metric.
VM runtime	`vm_uptime_seconds`	Active workspace VM time. Drives credits. Still-running sessions are included.
Tokens	`tokens_in` / `tokens_out` / `total_tokens`	LLM tokens consumed. Shown for insight; not billed.
Tasks / Agent runs	`task_count` / `agent_run_count`	How many tasks were active and how many agent turns ran in the window.

In Explore, the metric selector switches between Credits, Runtime, and Tokens. Tokens are not attributable per VM size, so the Tokens metric is unavailable when grouping by VM size.

Group-bys

Explore pivots the data along one dimension at a time. Group by:

Group by	Pivots on
Project	The project each task belongs to (default).
Agent	The coding agent (Codex, Claude Code, OpenCode, Cursor).
Model	The model the agent ran, scoped to its agent.
VM size	The runtime class (small, medium, large, xl). Carries runtime and credits only — tokens have no per-VM-size attribution.
Status	The task display status (running, awaiting_input, awaiting_ci, merged, unmerged, errored, archived).
None	No pivot — shows the per-task ledger instead of a group table.

The grouped table is interactive: clicking a row drills down one level (project → agent → model, or status/VM-size → the filtered ledger), narrowing the filters as it goes.

Per-task ledger

Group by None to see the paginated task ledger — one row per task with its credits, runtime, tokens, status, project, agent, model, PR number, and hardware. Sort by any of:

Sort field	Column
`name`	Task name
`started_at`	Started
`credits`	Credits
`runtime`	Runtime
`tokens`	Tokens

The ledger pages 25 rows at a time and defaults to sorting by credits, descending.

Date range and filters

A date-range control offers presets (7d, 30d, 90d, this billing cycle) and a custom range up to 366 days. Multi-select filters narrow by project, agent, model, status, and VM size. Filters and the active range apply to both tabs.

Export

In Explore, Export CSV downloads the current view. When grouped, it exports the complete group breakdown; with no grouping, it fetches every matching page of the task ledger, not just the visible page.

Open Settings → Usage, choose Overview or Explore, then set the range, group-by, and filters.

aether usage --from 2026-06-01 --to 2026-06-30 --group-by model

See aether usage for all flags.

Getting Started

Using Aether

Overview vs Explore

Metrics

Group-bys

Per-task ledger

Date range and filters

Export

​Overview vs Explore

​Metrics

​Group-bys

​Per-task ledger

​Date range and filters

​Export

Overview vs Explore

Metrics

Group-bys

Per-task ledger

Date range and filters

Export