PDF form filling
for AI agents.

# Shell + jq · replace the key and PDF path
API=https://pagewrite.dev/api/forms
KEY=cnk_…
form_id=$(curl -fsS "$API" -H "X-API-Key: $KEY" \
  -F "pdf=@housing_form.pdf" | jq -er '.form_id')

# Inspect the actual field IDs, labels and constraints
curl -fsS "$API/$form_id/fields" -H "X-API-Key: $KEY"
# Create values.json: {"values":{"FIELD_ID_FROM_SCHEMA":"VALUE"}}
curl -fsS "$API/$form_id/fill" -H "X-API-Key: $KEY" \
  -H "Content-Type: application/json" --data-binary @values.json

# Check the returned errors and validation before downloading
curl -fsS "$API/$form_id/download" -H "X-API-Key: $KEY" -o filled.pdf

# Illustrative tool calls for your MCP agent
form = analyze_form(pdf_path="/absolute/path/housing_form.pdf")
form_id = form["form_id"]
schema = list_fields(form_id=form_id)

# Choose field_id and value from the schema and user input
result = fill_fields(form_id=form_id, values={field_id: value})
# Inspect result.errors, warnings and validation; fix any rejects

download_filled_pdf(
  form_id=form_id, save_to="/absolute/path/filled.pdf"
)

// Example 200 response excerpt: one value rejected
{
  "filled_count": 0,
  "errors": [{
    "field_name": "field_015",
    "code": "value_too_long",
    "message": "Value has 6 characters but the field has only 5 character cells — shorten the value and re-send this field.",
    "expected_cells": 5, "provided_length": 6
  }],
  "warnings": [],
  "pdf_updated": false
}
// Valid values can succeed in the same request; check every error