Create your account
Go to sellag.vercel.app and sign up. Once you’re in, connect or generate a wallet address — this is where your earnings land. You can fund it later; you don’t need a balance to publish.
Create a new dataset listing
Navigate to Dashboard → Listings → New → Dataset. You’ll land on the listing creation form. Choose a storage mode before you proceed:
Inline
Best for small structured datasets. Content is stored in the catalogue record and returned directly in the response body.
Asset-backed
Best for large files. Content lives in object storage and buyers receive a short-lived presigned download URL.
Upload your file and fill in metadata
Upload your dataset file, then complete the metadata fields. Every field you fill in accurately improves your datacard and helps agents find and trust your listing.
Name
Name
Write a descriptive, keyword-rich title. Agents filter on keywords, so include real column names and the primary use case.
Description
Description
Explain what the data contains, how it was collected, and what tasks it suits. Agents read this before trialing.
Schema
Schema
List the fields, types, and any enum values. Agents with
required_fields set in their SKILLS.md check schema before buying.License
License
State the license type and provenance plainly. Ambiguous licensing causes agents to skip a listing. A clear
licenseType widens your buyer pool.Price
Price
Set
priceUSDC — the amount an agent pays per access. Use 0 for a free listing. For paid listings, see the pricing guidance below.Run analysis to build your datacard
Save the listing as a draft, then click Run Analysis. Sella’s ADC pipeline processes your dataset in three passes:The pipeline produces a
1
Structural analysis
Detects schema, formats, and completeness — the objective shape of your data.
2
Quality judging
Scores quality signals including task fitness and synthetic probability.
3
Trial generation
Builds the preview payload agents see when they call
try_dataset — a faithful sample, never your full content.dataCard on your listing with structural, judge, and trial sections. Agents read this card before deciding to buy, so a strong datacard is your best sales tool.While analysis runs, your listing is in
process state and invisible to agents. It becomes visible only once it reaches live or manual_review.Review your qualityStatus
Once analysis finishes, open your listing and check
qualityStatus. Aim for passed — that signals a green datacard and earns better placement in agent search results.If you don’t reach
passed, improve your metadata or data quality and re-run analysis. You can trigger a re-score at any time without re-uploading the file.Go live and start earning
Once
qualityStatus is passed, your listing moves to live automatically. From this point:- Agents can discover your listing via
list_datasetsand keyword search. - Agents can trial it with
try_datasetto inspect the sample before committing. - Agents can purchase it, triggering an x402 payment that settles to your wallet in USDC.
Next steps
Quality & datacards
Understand the signals behind your listing’s grade and learn how to optimize for a
passed status.Pricing & earnings
Choose the right tier, understand the 3% fee split, and set up withdrawals from your earnings wallet.

