Open-source data for humans and agents.
Esker is a registry for public data. Engineers and AI agents pull versioned, verifiable datasets — from any language, on any runtime.
Authoring
Publish once. Pull from anything.
Publish
$ esker push us.sec.companies esker-dev/us.sec.companies@1.0.0 10,348 records · 0.5s · output/us.sec.companies.parquet pushed esker-dev/us.sec.companies@1.0.0Consume
import esker esker.add("esker-dev/us.sec.companies") apple = esker.get_one( "us.sec.companies", where={"ticker": "AAPL"},)Live
What an Esker dataset looks like
esker/
ai.models.registry@1.4.0
model_idstring
providerstring
context_tokensinteger
modalityenum
releaseddate
claude-opus-4.5
anthropic
200000
text
2025-11-24
gpt-5
openai
400000
multimodal
2025-08-07
gemini-3-pro
google
1000000
multimodal
2025-12-01
claude-sonnet-4.5
anthropic
200000
text
2025-09-29
llama-4-maverick
meta
1000000
multimodal
2025-04-05
grok-4
xai
256000
multimodal
2025-07-09
mistral-large-3
mistral
256000
text
2025-10-15
1,284 records·weekly·CC-BY-4.0
TrendingAll datasets →