Quick overview
This workflow watches a Google Drive folder for new contracts, uses Anthropic Claude to extract and assess clauses against your legal playbook and hard limits, then aggregates a verdict, optionally pauses for a human decision, logs the results to Google Sheets, and sends notifications via a webhook.
How it works
- Triggers when a new file is created in a specific Google Drive folder (or runs manually with a provided file ID).
- Downloads the contract from Google Drive, converts it to Base64, and sends it to the Anthropic Messages API to extract a structured list of clauses.
- Splits the extracted clauses into small batches and uses Anthropic Claude to assess each clause against your playbook, producing severity, explanations, and suggested redlines.
- Aggregates all clause findings into a single review, deterministically sets a verdict (with hard-limit breaches forcing a “DO_NOT_SIGN”), and decides whether human review is required.
- If escalation is needed, posts decision links to your webhook endpoint and waits for a reviewer to proceed, request redlining, or reject; otherwise it auto-clears the contract.
- Builds a severity-ordered report, appends the full review record to a Google Sheets log, and posts a completion notice to the same webhook.
Setup
- Connect Google Drive credentials and set the folder to watch in the trigger (or set a test file ID for manual runs).
- Add Anthropic access for both extraction and assessment: an HTTP Header Auth credential with
x-api-key for the Anthropic Messages API request, and an Anthropic credential for the Claude assessment model.
- Update the configuration values for
ourSide, playbook, hardLimits, and (optionally) the Anthropic model name and clause batch size.
- Connect Google Sheets credentials and select the target spreadsheet and sheet tab used as the review log.
- Set
notifyWebhookUrl to an endpoint (for example, Slack/Teams incoming webhook) that can receive the reviewer decision message and completion notice.