Skip to main content
Discovery on Sella costs nothing. Before you evaluate a single dataset or call a single API, you can browse the entire catalogue — filtered by modality, format, price, and the exact column names your pipeline needs. Use discovery aggressively to cut the candidate list down to a handful of genuinely relevant results before you commit a cent.

​
Search the dataset catalogue

list_datasets and search_datasets accept identical filter parameters. Both return titles, descriptions, modality, format, tier, price, and dataset IDs — never paid content. Use list_datasets to browse broadly and search_datasets when you have a specific query in mind.
The REST endpoint is read-only and requires no authentication — useful for quick catalogue browsing outside of an agent loop.

​
Filter parameters

​
query
string
Free-text search across title, description, and tags.
​
category
string
Restrict results to a single category.
​
tier
string
One of free, standard, premium, or enterprise.
​
modality
string
structured for tabular data or document for text and PDF content.
​
format
string
csv, tsv, json, jsonl, parquet, txt, md, or pdf.
​
required_fields
string[]
Return only datasets whose schema contains every column in this list. This is the single most effective filter — it eliminates datasets you would reject anyway after paying for a trial.
​
max_price_usdc
number
Cap the per-access price in USDC. Set to 0 to see only free datasets.
​
sort_by
string
default:"newest"
Sort order: newest, popular, price_asc, or price_desc.
​
limit
number
default:"20"
Number of results to return, between 1 and 100.
Search free first. Sella mirrors open datasets from Hugging Face and Kaggle alongside premium listings. Filter with max_price_usdc: 0 — an open dataset often does the job and saves your budget for situations where quality genuinely demands a paid source.

​
Search the API Data Market

list_market_apis searches 1,400+ aggregated third-party providers spanning image and video generation, web search and scraping, social data, crypto and finance feeds, email, inference, and more.
Codes A1–A22 map to verticals — for example: A1 image generation, A3 video, A7 web search, A8 scraping, A9 social data, A10 email, A14 crypto, A16 finance, A19 inference. Pass the relevant code to narrow results to a specific domain.
B1 pay-per-call API · B2 data product · B3 inference/playground · B4 workflow · B5 infra/dev-tool · B6 physical commerce. Use B1 for standard per-request APIs.
chain (e.g. base, solana, stellar) restricts results to providers that settle on a specific network. verified_only hides unreviewed listings. max_price_usdc caps the per-call price. limit controls page size, defaulting to 25 with a maximum of 50.
Call list_market_apis at the start of each session rather than relying on cached results — provider availability and pricing change frequently. For full execution instructions, see API Data Market.

​
Discovery strategy

Lead with required_fields

Verifying schema up front eliminates irrelevant results before you spend any evaluation credits. Pair it with max_price_usdc to bound your budget in a single query.

Sort by popular

Purchase popularity is the fastest quality signal in the catalogue — what other agents repeatedly bought tends to be reliable. Start here before sorting by price.
Once you have a shortlist, move to evaluate to inspect schema coverage, sample rows, and quality datacards before you buy.