How we meter usage

Effective 2 November 2026

1. Agent Action. An "Agent Action" means any discrete step an agent takes to complete a request, including each tool call or model invocation an agent performs to retrieve information, interact with data systems, or produce output. Each such step is counted as one (1) Agent Action. Confidence-scoring evaluations are billed as Agent Actions at the same unit price. Agent Actions are billed independently of, and in addition to, any Page Parse, Low Priority Parsing, Deep Transformation, Query Token, or LLM Pass-Through fees.
2. Page Parse (Advanced). "Page Parse (Advanced)" means advanced document extraction (including layout analysis and optical character recognition), billed per page of each processed document. This charge is independent of, and does not count as, an Agent Action.
3. Page Parse (Text and Table). "Page Parse (Text and Table)" means standard text-and-table document extraction, billed per page of each processed document not processed as a Page Parse (Advanced) page. This charge is independent of, and does not count as, an Agent Action.
4. Low Priority Parsing. "Low Priority Parsing" means a Page Parse (Advanced) or Page Parse (Text and Table), run on lower-priority computing capacity when Customer chooses it for a document or data source. Pages are processed as capacity allows and take longer than standard parsing; no processing time is promised. It is billed per page, as Low Priority Parsing (Advanced) or (Text and Table), instead of the matching Page Parse charge, and is independent of Agent Actions.
5. Deep Transformation. "Deep Transformation" means Meibel's structured-data extraction service, billed on two independent units: (a) Deep Transform Pages, billed per page of source document processed by the service; and (b) Deep Transform Fields, billed per structured data field extracted by the service. Deep Transformation fees are independent of, and in addition to, Page Parse and Agent Action fees.
6. Ingest Tokens. "Ingest Tokens" means the source-content tokens ingested and indexed into Recall (the Platform's index and storage for Customer's content), billed per token of source content processed. Source content is metered once per document at the time of ingestion.
7. Query Tokens. Meibel operates and prices the models used by its query features. Usage is metered by token and billed separately for input and output tokens, and for TAG and RAG queries. Query Tokens are consumed by these Meibel-operated features and are distinct from, and billed separately from, LLM Vendor Fees (which apply to third-party models Customer selects; see Usage and Metering Schedule §4).
8. Recall Storage. "Recall Storage" means the volume of Customer Data stored within the Platform, metered in gigabyte-hours ("GB-hours"). Meibel measures the amount of Customer Data stored (in gigabytes) on at least an hourly basis and sums those measurements across each hour of the applicable billing period. Accordingly, one gigabyte stored continuously for one hour equals one GB-hour, and one gigabyte stored continuously for a full 730-hour month equals 730 GB-hours. Recall Storage is billed at the per-GB-hour rate in the Price List, after deducting the included GB-hour allotment for the applicable billing period (if any).
9. LLM Pass-Through. "LLM Pass-Through" means third-party model usage through Meibel's own provider accounts ("LLM Vendor Fees", Usage and Metering Schedule §4), charged as stated in the Price List.

Prices for each item are on the pricing page, or in the Order Form for customers who have one. Storage may be priced per GB-month (730 GB-hours) and tokens per million, but they are metered as described above.