AIData.Broker

Sourced prices, market rates, buyers, licensing deals and court rulings for AI training data, plus a valuation from public prices. Free, read-only, no key.

Hosted MCP Server

npx add-mcp 'https://aidata.broker/mcp'

Installs into Claude Code, Codex, Cursor and more

Documentation

The AI data market, inside your AI assistant.

Connect Claude, ChatGPT, Cursor or any assistant that speaks MCP to this site's research: 900 published prices, 51 market rates, 25 buyers and their terms, 95 licensing deals, 68 lawsuits and rulings, and the terms of 18 business systems. Every answer names its source and the page to cite. Free, read-only, no account.

Server address

https://aidata.broker/mcp

How to connect

Menu names move between versions of each app; the address is all any of them needs.

  • Claude (web and desktop) Settings, then Connectors, then Add custom connector. Give it a name, paste the address and save. In a chat, switch it on from the tools menu.
  • ChatGPT Settings, then Apps and Connectors, then Advanced settings: turn on Developer mode. Create a connector, paste the address, choose no authentication, and switch it on in a chat.
  • Claude Code Run this once in a terminal:
    claude mcp add --transport http aidata-broker https://aidata.broker/mcp
    
  • Cursor, VS Code and other clients Add it to the client's MCP settings as a remote (HTTP) server, for example:
    {
      "mcpServers": {
        "aidata-broker": {
          "url": "https://aidata.broker/mcp"
        }
      }
    }
    

What it answers

Market rates get_market_rates

What does this kind of data go for per hour, image, task or company, on each side of the market?

Price evidence search_prices

Every published price behind those rates, with the source's own words.

Value a holding value_data

What do public prices imply for this footage, these recordings, these records?

Buyers and their terms find_buyers

Who buys this kind of data, what they pay, how and when, and whether they pay introducers.

Licensing deals search_deals

Which licences of data to AI companies have been disclosed, for how much and how long.

Lawsuits and rulings search_cases

What courts and regulators have decided about training AI on data.

Software terms check_software_terms

Can a company sell what it holds in Slack, Salesforce, Zendesk and 15 more systems?

Open datasets list_datasets

Download every table as CSV or JSON under CC BY 4.0, with its DOI.

Questions to try

  • What do AI labs pay per hour of licensed video, and what do the people who shoot it get?
  • Our startup is closing. Who buys company records like Slack, email and tickets, and on what terms?
  • Can we sell our Zendesk tickets for AI training, or do Zendesk's terms get in the way?
  • What would 2,000 hours of call-centre recordings in Spanish be worth to a buyer?
  • Which publishers have licensed their archives to OpenAI, for how much and for how long?
  • What did the court decide in Bartz v. Anthropic, and what did it cost per book?

The same data is on this site's pages, starting with the market rates and the price evidence, and downloads as CSV and JSON from the datasets page. To value a holding in full, measure it on your own computer or offer it to buyers, use the valuation.

Questions about the server

What is the AIData.Broker MCP server?

An address, https://aidata.broker/mcp, that an AI assistant can connect to and query. It answers from the same published prices, rates, buyers, deals and rulings as this site, each with the source that stated it.

Does it cost anything or need an account?

No. It is free, needs no account and no key, and only reads: it cannot change anything or act on your behalf.

Where do the figures come from?

Every figure is one somebody else published: a court filing, a buyer's own page, a catalogue or named press. A valuation is the one thing it computes, and it says how, from those published prices. It is never an offer.

Can I reuse what it returns?

Yes, under CC BY 4.0: credit AIData.Broker, link the page each answer names, and keep the original source of each figure. The same data downloads as CSV and JSON from the datasets page.

What is recorded when my assistant calls it?

Which tool was called and the words it searched for are counted the way this site counts a page view: no account, no cookie, and no record of who asked.