Subprocessors

Runs the Claude models that read your uploaded contracts and remittance documents, hosted inside AWS. Our default provider: the model is the same one Anthropic runs, but the processing happens on AWS infrastructure.

Receives: Text extracted from uploaded documents, sent for processing and returned. AWS does not store it, does not use it to train models, and does not pass it to the model's vendor — when a model runs on Bedrock, Anthropic does not receive your content.

Reads uploaded contracts and remittance documents and extracts their terms into a structured form. Used directly when the AWS-hosted route is unavailable.

Receives: Text extracted from uploaded documents, sent for processing and returned. Not retained for training.

Alternative model provider for document reading and term extraction, used where it performs better on a document type.

Receives: Text extracted from uploaded documents, sent for processing and returned. Not retained for training.

Alternative model provider for document reading and term extraction, used where it performs better on a document type.

Receives: Text extracted from uploaded documents, sent for processing and returned. Not retained for training.

Database, authentication, and private file storage. The system of record for everything in your account.

Receives: Account credentials, organisation profiles, uploaded documents, findings, workpapers, disputes, ledger records, and the activity trail.

Application hosting and content delivery. Serves the site and runs the application's server-side code.

Receives: Request data in transit, including IP address and request headers. Uploaded documents pass through in transit but are not stored here.

Sends transactional email — results notifications, security alerts, password resets, and account email.

Receives: Recipient email address, and the content of the message we send you. Findings and documents are not attached to email.

Does not receive uploaded documents or their contents.

Error monitoring. Receives a diagnostic report when something in the application fails, so it can be fixed.

Receives: Error details, stack traces, and limited request context. Configured to avoid capturing document contents.

Does not receive uploaded documents or their contents.

Product analytics. Shows us which parts of the marketing site and application people use, so we can prioritise fixes and improvements.

Receives: On the marketing site: page views and on-page interactions, counted without a cookie or any identifier that persists between visits. In the application, only while the product's own cookie card allows it — after the person accepts where the law requires asking first, and until the person turns analytics off elsewhere: page views, a short list of named product events (an upload started or completed, a finding opened, a claim recorded), and an internal user ID, organisation ID and name, and role, used to tie usage to an account. One cookie, kept for at most six months and removed the moment analytics is turned off. No uploaded documents, findings, workpapers, buyer names, or dollar figures are captured, and screen or session recording is off.

Does not receive uploaded documents or their contents.