​

Set up Dex, the document data extractor

Install the Document Data Extractor template, connect one channel, optionally connect an output destination, and define your schemas. Dex pulls structured fields from documents — and never fabricates.

Dex is a data-extraction agent built as an Agent Hub template. It pulls the exact fields you need from documents, emails, and forms and returns them in a clean, consistent format. Missing or ambiguous fields are flagged, not guessed.

​

Before you install

• A running Nora instance with dashboard access.
• The document types you process and the fields you need from each.

​

Step 1 — Install from Agent Hub

Open Agent Hub, find Document Data Extractor Claw, and click Install. See Publish and install Agent Hub templates.

​

Step 2 — Connect a channel

Dex needs one channel to send you extractions and flags (Channels tab). Or use Telegram / Slack / Discord / email — see Configure channels.

​

Step 3 — (Optional) Connect an output destination

Dex extracts fine from documents you paste — no integration required. To write results straight out, connect one in the Integrations tab:

​

Step 4 — Run the bootstrap

Send any first message. Dex introduces itself, offers to rename, gets your channel connected, then defines a field schema per document type and your preferred output format (table / JSON / CSV / list).

​

What Dex will and won’t do

• Never fabricates — ambiguous fields come back raw and flagged.
• Stable schemas — the same document type produces the same output shape every time.

​

Troubleshooting