moraGet started
Introduction
What Mora is for a developer or an agent: one search over private and free training data, and a fixed way to list and deliver a dataset.
Mora is a market for AI training data. Companies license the records they already have (tickets, code, claims, recordings) to AI labs, and neither side is ever named to the other.
For a developer, Mora is three things:
- One search over two shelves. Private datasets licensed through Mora, and free public datasets from Hugging Face and Kaggle with what their license allows.
- An MCP server. Your agent (Claude, ChatGPT, Cursor) searches, tests samples, lists your company's records and follows your deals.
- A dataset format. Standards made of parts, so data arrives in the same shape from every company.
Who uses it
| You are | You use Mora to |
|---|---|
| Building a model, no budget | Find free datasets whose license allows your use, and look inside them before downloading |
| A lab or a team training a model | Find private data, test a sample Mora drew at random, make an offer |
| A company with records | Have your agent build the dataset, take people out of it, and list it |
Start here
- Quick start: connect an agent in two minutes.
- Tools reference: every MCP tool with its parameters.
- REST API: the same data as plain JSON.
- Sending a dataset: the four calls a provider's agent makes.
What an agent never does
An agent can search, describe, prepare and send records. It never pays, signs, sends an offer or authorizes a tool. For those it returns a mora.market page and a person does it.