> ## Documentation Index
> Fetch the complete documentation index at: https://data-docs.orbbit.co/llms.txt
> Use this file to discover all available pages before exploring further.

# Data MCP

> Connect Orbbit Data to an AI assistant through MCP.

Orbbit Data has its own connector, so you can ask an AI assistant for commodity and branded-food data in plain words. The assistant picks the right tool, calls it with your account, and answers from the result.

It uses MCP (Model Context Protocol), the standard way AI tools plug into outside sources. This is a different connector from the [Docs MCP](https://data-docs.orbbit.co/for-agents/docs-mcp.md): that one lets an assistant read these pages, while this one fetches Orbbit's data.

Connector address: `https://data-api.orbbit.co/mcp`

Your console's **MCP** page shows the same address with a copy button.

## Claude (web, desktop, mobile)

See [Use With Claude](https://data-docs.orbbit.co/for-agents/use-with-claude.md) for step-by-step instructions with a screenshot.

## Claude Code

> [tab] Sign in
>
> Run this, then sign in to Orbbit and click **Allow** when your browser opens:
>
> ```bash theme={null}
> claude mcp add --transport http orbbit-data https://data-api.orbbit.co/mcp
> ```

> [tab] API key
>
> Connect with one of your [API keys](https://data-docs.orbbit.co/authentication.md) instead of signing in:
>
> ```bash theme={null}
> claude mcp add --transport http orbbit-data https://data-api.orbbit.co/mcp \
>   --header "Authorization: Bearer orbbit_data_YOUR_KEY"
> ```
>
> The assistant only sees the tools whose endpoints that key is scoped to. Revoking the key cuts the connection off.

## What the assistant can do

The connector gives the assistant five tools, one per endpoint. Each tool takes the same inputs as its endpoint, and its title is the endpoint's name in the [API Reference](https://data-docs.orbbit.co/api-reference/introduction.md). If you connect with an API key, the assistant only sees the tools that key is scoped to.

| Tool | Shown as | What it does | Endpoint | Price |
| - | - | - | - | - |
| `list_commodities` | List Commodities | List the price series we track, like butter or corn, narrowed by category and commodity, attributes, market and series. | `GET /v1/commodities` | Free |
| `get_commodity_metric_values` | Get Commodity Metric Values | Get the values over time for the series you choose, each described in the same four layers. | `GET /v1/commodities/metric-values` | \$0.01 |
| `list_branded_food_categories` | List Branded Food Categories | List every food category you can search within. | `GET /v1/branded-foods/categories` | Free |
| `search_branded_foods` | Search Branded Foods | Search for packaged food products by name, brand, category, barcode, or ingredient. | `GET /v1/branded-foods` | \$0.01 |
| `get_branded_food` | Get a Branded Food | Get everything we have on one packaged food product. | `GET /v1/branded-foods/{id}` | \$0.01 |

Every tool only reads data. A typical question about prices takes two tools: `list_commodities` to find the series, then `get_commodity_metric_values` to get its values.

Two things differ from calling the API yourself, to keep answers small enough for the assistant to read:

* When the assistant doesn't set `limit`, `list_commodities` and `search_branded_foods` return 10 rows per page.
* When the assistant doesn't set `start_date`, `get_commodity_metric_values` returns each series' recent stretch, the same as the API does. See [Commodities](https://data-docs.orbbit.co/market-data/commodities.md#when-you-leave-out-the-dates).

Every tool call goes through the same API as a direct request with your key, so the answer matches what you'd get from the API yourself. Each call to a paid tool is charged like any other. See [Pricing & Billing](https://data-docs.orbbit.co/pricing.md).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.