Managing Cloud Usage

Managing cloud usage

GET /v1/usage is the only endpoint for quota and billing visibility. It returns aggregate counters for the queried period plus the remaining quota for the project.

Endpoint

GET https://api.kreuzberg.dev/v1/usage
Authorization: Bearer $KREUZBERG_API_KEY

Query parameters

Both are optional. Omit both for the current calendar month.

Response (200)

{
  "period_start": "2026-05-01",
  "period_end": "2026-06-01",
  "total_pages": 5432,
  "total_documents": 87,
  "total_failed": 2,
  "quota_limit": 100000,
  "quota_remaining": 94568,
  "by_mime_type": {
    "application/pdf": { "documents": 65, "pages": 3200, "failed": 1 },
    "image/png":       { "documents": 15, "pages": 1800, "failed": 0 },
    "text/plain":      { "documents":  7, "pages":  432, "failed": 1 }
  }
}

Reading the response

• total_pages — pages billed in the period. The unit of cost.
• total_documents — files submitted, regardless of page count.
• total_failed — extractions that ended in failed status. Failed jobs do not consume quota.
• quota_limit / quota_remaining — total and remaining pages on the current plan.
• by_mime_type — per-MIME breakdown. Useful for identifying which document types drive cost.

Examples

Current-month usage

curl -s https://api.kreuzberg.dev/v1/usage \
  -H "Authorization: Bearer $KREUZBERG_API_KEY" | jq .

Specific date range

curl -s "https://api.kreuzberg.dev/v1/usage?start=2026-01-01&end=2026-02-01" \
  -H "Authorization: Bearer $KREUZBERG_API_KEY" | jq .

Quota remaining as a percentage

curl -s https://api.kreuzberg.dev/v1/usage \
  -H "Authorization: Bearer $KREUZBERG_API_KEY" \
  | jq '.quota_remaining * 100 / .quota_limit'

When to report usage to the user

Pull usage proactively when:

• A batch job submits more than ~100 documents — report quota_remaining after submit so the user can see the impact.
• The user asks "how much have I used?" or any quota-shaped question.
• A 429 response includes a quota-exhausted error — surface the usage shape so the user can decide whether to upgrade.
• After a long-running crawl finishes, since page count is hard to estimate up front.

Don't report usage on every routine extraction — it's noise.

Errors

Source: hashgraph-online/awesome-codex-plugins → plugins/kreuzberg-dev/plugins/plugins/kreuzberg-cloud/skills/managing-cloud-usage/SKILL.md