Extraction Agent Quick Start Guide

Getting Started

Sign up for a free Cloudsquid account (no credit card required) and follow these steps to start extracting and automating your document workflows.

1. Create a New Extraction Agent

All processing in Cloudsquid starts with an Extraction Agent.
  1. From the dashboard, click New Extraction Agent.
  2. Give it a name and click Create. Recording2025 08 13at15 21 43 Gi
  3. You’ll be taken into the extraction table interface with the Document Co-Pilot on the left and your table in the center.

2. Upload Files

The best way to start is by adding files to your table:
  • Drag and drop files into the center of the table, or
  • Click Add Files to select them from your computer. Addfiles Gi
Advanced options:
  • API Upload – Send files directly via the API.
  • Workflow Builder Connectors – Add files automatically from integrated systems (covered later).
Once uploaded, files will appear in your table with the status Waiting for run.

3. Define Your Extraction Schema with the Assistant

On the left-hand side, the Document Co-Pilot is ready to help you:
  1. Describe the task or job you want to perform on your uploaded files.
  2. The assistant reads your files and your description, then proposes extraction tasks (columns) for your table.
  3. Click Apply Schema to add these tasks to the table. Recording2025 08 13at13 59 35 Gi

4. Run Your First Extraction

  1. Select one or more files in the table.
  2. Click Run AI to process them.
  3. Extracted data appears directly in the table. Recording2025 08 13at15 29 59 Gi

5. Review and Refine

You can check and improve your extraction results in several ways:
  • Use the File Viewer – View the original file alongside the table data. Recording2025 08 13at14 06 36 Gi
  • Iterate with the Assistant – Ask it to add, remove, or adjust tasks. Recording2025 08 13at14 08 36 Gi
  • Manual Adjustments
    • Click Add Extraction Tasks to create new columns. Recording2025 08 13at14 11 45 Gi
    • Use the 3-dot menu on a column to edit an existing task. Recording2025 08 13at14 13 31 Gi
  • Run AI after edits – Select the range of files you want to re-extract data from after making adjustments to the columns Recording2025 08 13at15 25 21 Gi

6. Additional Features & Tools

  • System Prompt – In the settings, adjust the high-level guidance you want the AI to follow across all extractions. Recording2025 08 13at15 32 57 Gi
  • Bounding Boxes (New) – Enable this to highlight the exact location in the source document for any extracted value.
    • Useful for auditability, but may be slower for AI runs and is still in beta. Screenshot2025 08 13at15 35 56 Pn
  • Bulk Actions – Select multiple files, right-click, and choose Export or Delete. Recording2025 08 13at15 38 16 Gi
  • Long Array Mode (Lists only) – Enable for very large lists to avoid context window limits. Recording2025 08 13at15 39 28 Gi

7. Exporting Your Data

  • CSV or Excel – Download your entire table or selected rows by selecting files and using the bulk actions.
  • Webhook Delivery – Send data to an endpoint of your choice for real-time integration with other systems. Recording2025 08 13at15 42 11 Gi

8. Workflow Builder (Early Access)

The Workflow Builder is your automation hub in Cloudsquid, connecting your extraction tables to thousands of external tools and services. You can:
  • Import documents from sources like SharePoint, Amazon S3, and databases.
  • Send extracted data to destinations like Postgres, Snowflake, Salesforce, or Slack.
  • Automate multi-step processes after extraction, such as matching invoice data to transaction records in another system.
This feature is ideal for building end-to-end, automated workflows—covering both the ingestion of documents into Cloudsquid and the delivery of structured data to other systems.
Note: A dedicated Workflow Builder Quickstart Guide is available on the side bar

Next Steps

You now have the basics for:
  1. Creating an Extraction Agent
  2. Uploading and processing files
  3. Reviewing and refining results
For advanced topics—API integration, complex schema design, or Workflow Builder deep dives—refer to the detailed sections of the Cloudsquid documentation or reach out to the team via chat / email