Usage in the console

The Usage page shows what your organization ran and spent: totals for a time range, a runs-per-day chart, spend by model, the quota for the current period, and a request log with one row per model call. Use it to check what a run cost, to see how close you are to a spend limit, or to find a failed run. Open it at console.opentype.dev/usage. It needs the usage_read scope, which every role holds. Every figure comes from the same usage routes your code can call, listed at the end of this page.

Usage page with run, token, spend and error-rate totals, a runs-per-day chart, and period spend and token quotas — Usage with sample figures

Pick a range

The range tabs in the header are 24h, 7d and 30d; 30d is the default. Ranges count whole UTC days up to now:

Tab	Covers
24h	Today, from 00:00 UTC until now
7d	The last 6 whole UTC days and today so far
30d	The last 29 whole UTC days and today so far

The line under the title states the exact window, for example last 7d · 2026-09-18 to 2026-09-24 14:05 UTC · counted from settled accounting rows. The range applies to the totals, the chart, spend by model and the request log. It does not apply to Quota, this period, which always covers the current calendar month.

Totals

Five figures sit across the top.

Figure	What it counts	The line under it
Runs	Runs created in the range	How many completed, and how many are still in flight
Tokens in	Input tokens over those runs	provider-reported or estimated
Tokens out	Output tokens over those runs	Tokens in and out together
Spend	Settled spend plus what in-flight runs still hold, in US dollars	The settled part, and the part held
Error rate	Failed runs as a share of all runs, such as `1.5%`	How many runs failed

Amounts under a cent show four decimals, such as $0.0019, so small spend does not read as $0.00. A single decision run can cost less than

0.0001, which shows as `

0.0000`. See Models and pricing for the price per token.

Runs per day

Runs per day has one bar per UTC day in the range, scaled to the busiest day; its meta shows that peak, such as peak 214 · UTC. Hover a bar to see its date, its run count and its spend. Days with no runs are drawn flat rather than skipped. The last bar is today, labelled partial because the day is not over.

Spend by model

Spend by model groups the request log by model. It has one row today: Neon 1.1, served by OpenType.

Column	Meaning
Model	The model name, with who served it underneath
Attempts	Model calls in the log
Tokens	Input and output tokens together
Spend	What those calls cost

Its meta says what it is built from, such as from the last 200 attempts: it adds up the same rows as the request log, so in a busy range it covers the newest 200 calls rather than the whole range. The Spend total at the top always covers the whole range.

Quota, this period

Quota, this period shows where the organization stands against its limits for the current quota period, the calendar month in UTC. Its meta shows when the period ends, such as resets 2026-10-01.

Row	What it shows
Period spend	Spend this period, settled plus held, with the limit beside it (`of $50.00`), or no limit set
Period tokens	Tokens this period, with the limit beside it, or no limit set
Per-request ceiling	The most a single run may cost, at most $0.02. Each run holds this much while it is in flight
Held, not settled	What runs still in flight are holding

The panel reads the same figures the API checks when it admits a run, so this panel and a refusal never disagree. When a run would pass the period’s spend or token limit, POST /v1/runs refuses it with 429 organization_spend_quota_exhausted or 429 organization_token_quota_exhausted, before anything is read or charged. The limits reset at the start of the next period. See Spend limits and quotas. The same period spend appears in the sidebar as Spend this period, against the cap.

Request log

The Request log has one row per model call, newest first, from an append-only record: rows are added, never changed. It shows the newest 200 calls in the range; the count above it reads, for example, 37 of 200 attempts. A run usually makes one call; a run that was retried has one row per attempt.

Column	What it shows
Run · attempt	The run ID, with `attempt 0`, `attempt 1` and so on underneath
Model	`Neon 1.1`, with `OpenType` underneath
Status	`succeeded` or `failed`
Tokens in / out	Input and output tokens for this call
Cost	What this call cost
Created, UTC	When the call was made, such as `09-24 14:02:11`

Filter the log

The filter bar above the log narrows the rows already loaded:

Run ID: a search box (placeholder run_…) that matches any part of the run ID.
Status: Status: any, or a status from the rows loaded, such as failed.
Model: Model: any, or Neon 1.1.

Choose Apply to filter and Clear filters to reset. The range and filters stay in the page address, so you can bookmark or share a filtered view.

The log prints run IDs with hyphens, run_a4314b6c-c08f-4bd8-8140-99a613abeb44, while the Playground and POST /v1/runs print them without, run_a4314b6cc08f4bd8814099a613abeb44. Both name the same run, and GET /v1/usage/runs/{run_id} accepts both. To find a run in the log, search for run_ plus its first 8 characters, such as run_a4314b6c, which are the same in both forms.

With filters that match nothing, the log reads No attempt matches these filters: clear a filter or widen the range. With no calls in the range at all, it reads No attempt in this window.

A new organization

Until the organization has run anything in the range, the totals are replaced by No usage recorded yet: Usage is written when a run settles. Send a request from the Playground or with an API key and this page fills in. Choose Open the Playground to run one.

A run shows up in the totals as soon as it is created, and its cost once it settles, which for a decision run is when its answer comes back.

When figures are missing

The page never shows a figure it could not read as zero.

You see	What it means	What to do
The usage rollup could not be read, with a status and code	The totals could not be loaded. Figures are withheld rather than shown as zero	Reload. If it persists, look the code up in Errors
`[!] Daily usage unavailable: 503 database_unavailable`	One panel could not be loaded; the rest of the page still works	Reload in a moment
`[!] Quota unavailable: …` or `[!] The ledger unavailable: …`	The same, for the quota or the request log	Reload in a moment
`403 scope_denied`	Your credential lacks `usage_read`	Every role holds it; sign out and in again, or check the key’s scopes. See scope_denied

Read the same figures from code

Each panel is one API route, and all of them need usage_read. A key created with the Read only set on API keys has it.

Panel	Route
Totals	`GET /v1/usage?start_at=…&end_at=…`
Runs per day	`GET /v1/usage/daily?start_at=…&end_at=…` (at most 92 days)
Request log, Spend by model	`GET /v1/usage/ledger?start_at=…&end_at=…&limit=200`
Quota, this period	`GET /v1/quota`
One run’s cost and attempts	`GET /v1/usage/runs/{run_id}`

Windows are UTC timestamps in exactly the form 2026-09-01T00:00:00Z, sent together or not at all. Leaving both out means the current quota period.

# Where the organization stands this period
curl -sS https://api.opentype.dev/v1/quota \
  -H "Authorization: Bearer $OPENTYPE_API_KEY"

# Totals for September 2026
curl -sS "https://api.opentype.dev/v1/usage?start_at=2026-09-01T00:00:00Z&end_at=2026-10-01T00:00:00Z" \
  -H "Authorization: Bearer $OPENTYPE_API_KEY"

const auth = { Authorization: `Bearer ${process.env.OPENTYPE_API_KEY}` };

const quota = await fetch("https://api.opentype.dev/v1/quota", { headers: auth }).then((r) => r.json());
const spent = quota.consumed_spend.settled_micros + quota.consumed_spend.unsettled_micros;
const limit = quota.limits.period_spend_limit_micros; // null when no limit is set
console.log(`spent $${spent / 1e6}`, limit === null ? "no limit" : `of $${limit / 1e6}`);

const params = new URLSearchParams({ start_at: "2026-09-01T00:00:00Z", end_at: "2026-10-01T00:00:00Z" });
const usage = await fetch(`https://api.opentype.dev/v1/usage?${params}`, { headers: auth }).then((r) => r.json());
console.log(usage.runs.total, "runs,", usage.runs.failed, "failed");

import os

import requests

auth = {"Authorization": f"Bearer {os.environ['OPENTYPE_API_KEY']}"}

quota = requests.get("https://api.opentype.dev/v1/quota", headers=auth, timeout=30).json()
spent = quota["consumed_spend"]["settled_micros"] + quota["consumed_spend"]["unsettled_micros"]
limit = quota["limits"]["period_spend_limit_micros"]  # None when no limit is set
print(f"spent ${spent / 1e6}", "no limit" if limit is None else f"of ${limit / 1e6}")

usage = requests.get(
    "https://api.opentype.dev/v1/usage",
    params={"start_at": "2026-09-01T00:00:00Z", "end_at": "2026-10-01T00:00:00Z"},
    headers=auth,
    timeout=30,
).json()
print(usage["runs"]["total"], "runs,", usage["runs"]["failed"], "failed")

GET /v1/usage answers with the figures behind the totals. Money is integer micro-USD: 1,000,000 is $1.

{
  "organization_id": "org_example",
  "window": {"start_at": "2026-09-01T00:00:00Z", "end_at": "2026-10-01T00:00:00Z"},
  "runs": {"total": 3, "completed": 2, "failed": 0, "in_flight": 1},
  "tokens": {"input_tokens": 824, "output_tokens": 46, "total_tokens": 870},
  "spend": {"reserved_micros": 20038, "settled_micros": 38, "unsettled_micros": 20000}
}

Here one run is still in flight and holds

0.02 (`unsettled_micros: 20000`), so **Spend** reads `

0.0200` (settled plus held) until the run settles. Usage reporting covers every route and field.

Usage reporting - every usage route, its fields, and how to build your own reports.
Spend limits and quotas - the period limits, the per-request ceiling, and the 429 refusals.
Billing - the balance that each run’s cost is charged against.
Models and pricing - the price per token behind every Cost figure.
Request IDs - tie a row in the log to the request your code sent.

Overview

Workspace

Pick a range

Totals

Runs per day

Spend by model

Quota, this period

Request log

Filter the log

A new organization

When figures are missing

Read the same figures from code

​Pick a range

​Totals

​Runs per day

​Spend by model

​Quota, this period

​Request log

​Filter the log

​A new organization

​When figures are missing

​Read the same figures from code

​Related

Pick a range

Totals

Runs per day

Spend by model

Quota, this period

Request log

Filter the log

A new organization

When figures are missing

Read the same figures from code

Related