AI Agent & Evidence
The AI Agent helps you build UNTP-conformant credential content from your existing documents. Upload evidence, describe your facility or product, add conformity claims, and the AI generates a complete credentialSubject that passes Tier 2 schema validation.
The AI Agent is live for DFR (Digital Facility Record) credentials. DPP AI Agent is coming soon — the same workflow will apply to Digital Product Passports.
How It Works
The credential editor has four tabs: Form, JSON, AI Chat, and AI Agent. The AI Agent tab is where you provide context — documents, descriptions, and conformity claims — that the AI uses to generate credential content.
Upload documents + describe facility/product + add claims
→ AI generates credentialSubject
→ Validates against UNTP schema (up to 3 attempts)
→ Result saved to your draft
→ Refine further via AI Chat tab
Step-by-Step
1. Open the AI Agent Tab
Navigate to Toolkit > Credentials, find your draft credential, click Edit, then select the AI Agent tab.
2. Upload a Photo
Upload a facility image (DFR) or product image (DPP). This becomes a publicly accessible image link in the credential. Accepted formats: PNG, JPG, JPEG, WebP.
3. Write a Description
Describe the facility or product in your own words. The AI uses this as primary context. Be specific — mention what the facility does, what it produces, where it's located, or what makes the product notable.
Check "Allow AI to edit and enhance the description" if you want the AI to refine your text.
4. Upload Supporting Documents
Upload PDFs, Word documents, spreadsheets, or text files that contain relevant information. The AI extracts text from these files and uses it as context when generating the credential.
Supported formats: PDF, DOCX, XLSX, CSV, TXT (max 20MB per file).
Examples of useful supporting documents:
- Facility registration or operating permits
- Product specification sheets
- Environmental impact reports
- Operational capacity documentation
5. Add Conformity Claims
Conformity claims are the heart of the credential — they declare what standards, certifications, or regulations your facility or product complies with. Each claim becomes a separate entry in the credential's conformityClaim array.
Click "Add Claim" to create a new claim. For each claim you can provide:
Claim Description (required)
A short description of the conformity claim. Examples:
- "Facility is certified to ISO 14001:2015 Environmental Management"
- "Product meets EU Deforestation Regulation (EUDR) requirements"
- "Organic certification from Australian Certified Organic"
Evidence Document (PDF upload)
Upload the actual certificate, audit report, or test result as a PDF. The AI reads the document text and uses it to populate the claim's evidence fields in the credential.
For example, uploading an ISO 14001 certificate PDF lets the AI extract the certificate number, issuing body, validity dates, and scope — details you'd otherwise need to enter manually.
DCC Link (existing Digital Conformity Credential)
If a conformity assessment body has already issued a Digital Conformity Credential (DCC) for this claim, paste the DCC URL instead of (or in addition to) uploading a PDF.
DPP Kit will:
- Fetch the DCC from the URL (resolving through the Identity Resolver if needed)
- Verify its cryptographic signature
- Check whether the DCC references your credential's identifier
- Show a validation badge:
- Green — Valid signature and references your identifier
- Yellow — Valid signature but doesn't reference your identifier
- Red — Invalid signature or unreachable
The AI uses the verified DCC metadata to populate the conformity claim with accurate assessment details.
You can use both a PDF upload and a DCC link on the same claim. The PDF provides human-readable evidence while the DCC provides machine-verifiable proof. The AI incorporates context from both.
6. Run the AI Agent
Click "Generate with AI Agent". The AI:
- Reads all your uploaded documents and extracts their text
- Combines your description, claims, and document context
- Generates a complete
credentialSubjectconforming to the UNTP schema - Validates its output against the schema (retrying up to 3 times if needed)
- Saves the result to your draft credential
After generation, you're switched to the AI Chat tab where you can see the result and refine it conversationally.
Managing Claims
Existing vs. New Claims
After the AI generates your credential, any claims it created become "Saved" claims (shown with a lock badge). On subsequent visits to the AI Agent tab:
- Saved claims are shown in a read-only collapsed view with their description, evidence, and DCC link
- New claims can be added alongside existing ones
- Running the AI again preserves saved claims and only generates content for new ones
Deleting a Claim
To remove a claim, expand it and click "Delete this claim". This:
- Removes the claim from the credential
- Deletes any associated evidence files (so the AI won't re-create the claim from orphaned documents)
- Updates the draft in Directus immediately
Editing Existing Claims
Saved claims cannot be edited directly in the AI Agent tab. To modify a claim's content, use the AI Chat tab — describe what you want to change and the AI will propose updates in a diff view that you can accept or reject.
After Generation
Once the AI has generated your credential content:
- Review the result in the Form or JSON tab
- Refine using the AI Chat tab if needed
- Run Tier 2 Tests to confirm schema compliance
- Issue the credential when you're satisfied
Running the AI Agent again will replace the existing credential content (except for preserved claims). If you've made manual edits in the Form or JSON tab, they will be overwritten.