# Quickstart (/docs/quickstart)

quarry gives your assistant sources it can cite: reports, filings, research, and
data sold by the people who produced them. Your assistant adds quarry by address,
searches it for free, and gets passages from free datasets and from the ones you
pay for. Unsure of a word? See the [glossary](/docs/glossary).

## 1. Add quarry to your assistant [#add-quarry]

quarry is an MCP server. MCP is the standard way assistants plug in outside tools:
you give them an address.

quarry has two:

* The free address, `https://quarry.market/api/mcp`, to search quarry and read free
  datasets with no account. In Claude, it also asks you to connect your quarry
  account when you use a subscription.
* The sign-in address, `https://quarry.market/api/agent/mcp`, to use your monthly
  subscriptions from assistants that sign in first, like ChatGPT. Your assistant
  asks you to sign in to quarry once (or to create an account), and you click
  **Connect agent**.

In Claude, [add quarry in one click](https://claude.ai/customize/connectors?modal=add-custom-connector\&connectorName=quarry\&connectorUrl=https%3A%2F%2Fquarry.market%2Fapi%2Fmcp)
and choose Sign in when needed. For ChatGPT, see
[Chat assistants](/docs/connect/assistants). In Claude Code, run this command, then
`/mcp` inside Claude Code to pick quarry and sign in:

```bash
claude mcp add --transport http quarry https://quarry.market/api/agent/mcp
```

Other tools are in [Connect your tool](/docs/connect).

## 2. Ask a question that needs sources [#ask]

Turn quarry on in the chat (in Claude: the + button, then Connectors), then ask
something your assistant shouldn't answer from memory, for example "What did the
latest outlook reports say about small-business growth? Cite your sources." It
searches quarry for free and gets back the datasets that match, with their price
and, for the top results when the seller allows it, a short preview of the best
passage.

## 3. Get the passages [#get-the-passages]

To read a dataset, your assistant asks it the question and gets the passages that
answer it:

* Free datasets answer right away.
* Paid datasets use your monthly subscription. If you don't subscribe to that
  dataset yet, your assistant tells you its price and can prepare a Checkout link:
  see [Subscriptions and API keys](/docs/subscriptions).
* Or your assistant pays per question from a USDC wallet of its own: see
  [Give your assistant a wallet](/docs/wallet).

When the relevance check finds no passage that answers the question, nothing is
charged. Passages that answer only part of it still count.

## 4. Cite the sources [#cite]

Each passage comes with its file name, page, and the dataset's listing link. Your
assistant can quote them and cite the source, and you can open the listing to see
who published it.

## Building your own agent? [#building-your-own-agent]

* Load quarry's tools in your framework: [Frameworks and SDKs](/docs/connect/frameworks).
* Pay per question from a wallet, over MCP or plain HTTP: [Paying with x402](/docs/x402).
* Every tool and parameter: [MCP server](/docs/mcp).
* To check a wallet setup, `https://quarry.market/api/x402/demo` charges 0.001 USDC and
  returns a test answer.