MCP server

Use Rare Context in Claude, ChatGPT or Cursor

Connect Claude, ChatGPT, Cursor or another AI assistant to Rare Context, and it can look up power plants, data centers, substations, mines, water rights and powered land, who owns them, and the source of every fact. The server is at https://rarecontext.com/mcp: free, read-only, and no sign-in.

Add it

The address is the same in every client: https://rarecontext.com/mcp.

No connector? Paste our prompt into Claude or ChatGPT instead: the assistant then reads the same data through plain links, with nothing to install.

Claude (claude.ai, Claude Desktop and mobile)

Add to Claude

Or by hand:

Open Customize → Connectors and choose Add custom connector.
Name it Rare Context, paste https://rarecontext.com/mcp and add it. There is no sign-in.
In a chat, turn Rare Context on in the tools menu and ask a question.

Every plan can add it; the free plan has room for one custom connector. On Team and Enterprise, an owner adds it under Organization settings → Connectors, and members then connect.

Claude Code

claude mcp add --transport http rarecontext https://rarecontext.com/mcp

Add --scope user before the name to have it in every project.

ChatGPT

Go to chatgpt.com/plugins, press + and choose Add custom MCP server.
Name it Rare Context, paste https://rarecontext.com/mcp, choose No authentication, accept the warning and choose Create as a plugin.
In a chat, type @ and pick Rare Context.

On the web. On Business, Enterprise and Edu, an admin adds it.

Cursor

Add to Cursor

Or add it to .cursor/mcp.json in a project, or ~/.cursor/mcp.json for every project:

{
  "mcpServers": {
    "rarecontext": {
      "url": "https://rarecontext.com/mcp"
    }
  }
}

VS Code

Add to VS Code

Or add it to .vscode/mcp.json, or run MCP: Add Server from the command palette and choose HTTP:

{
  "servers": {
    "rarecontext": {
      "type": "http",
      "url": "https://rarecontext.com/mcp"
    }
  }
}

Any other MCP client

Add a remote server with the Streamable HTTP transport at https://rarecontext.com/mcp, with no authentication.

Try asking

Which are the largest data centers in the United States, and who owns them?
Find sites of 20 ha or more within 1 km of a 380 kV substation in Gelderland, outside Natura 2000.
Who owns the W A Parish power plant, and what else does that owner hold?
Brownfield sites in Leeds near a Northern Powergrid substation with 20 MVA of headroom or more.
What does Blackstone own in real assets, and who are its subsidiaries?

What it can do

Tool	What it does
`search`	Find Rare Context pages by name: assets (power plants, data centers, substations, mines, fields, water rights, grid connections), owners and investors, places and sources.
`search_sites`	Screen candidate sites for data centers, batteries or industry near grid capacity in the Netherlands (NL), Great Britain (GB) or Germany (DE): the filters of https://rarecontext.com/sites/search, with the same code.
`get_entity`	Read one asset, right or site: a power plant, data center, substation, grid connection, mine, oil or gas field, licence, water right or powered-land site.
`get_owner`	Read one owner, operator or investor: its portfolio of assets by class and country, its parent company and subsidiaries (the group, with its rolled-up portfolio on the parent's page), funds it manages, shares in listed asset owners, partners, dated events and its identifiers (LEI, company number), each fact with its source, licence and date.
`rankings`	The largest assets of a class by its measure (capacity, IT power, reserves, volume, area), worldwide or in one country, or its owners ranked as corporate groups.
`request_research`	The link where the person can ask Rare Context for what the pages do not publish: an owner, grid capacity, planning status, a listed figure, research on an asset or region, or an introduction to an owner.

Every answer is read from the pages of this site: it gives the page of each asset, owner or site it names, and each fact's source, licence and the date we last checked it. A fact a source does not let us republish is marked available on request, with the link to ask for it. The assistant can give you that link; request_research fills in the request form, and you send it yourself with your name and email.

Limits and privacy

Read-only: the tools cannot change anything.
No account or key. At most 120 requests a minute from one address; the count is kept under a hash of the IP address, which we never store. Claude and ChatGPT call from their own shared addresses and have a limit of their own.
The server keeps no record of what you ask. As for any page, we count the visits of AI assistants by page path and assistant only (see privacy).
The data keeps the licence of each source: credit the source named with a fact when you reuse it (data licence).

Support

A tool that fails or answers wrongly: email hello@rarecontext.com with the question you asked and the tool's answer. A person reads every message.

For developers

Streamable HTTP at https://rarecontext.com/mcp: MCP 2026-07-28 (stateless, server/discover) and the earlier revisions back to 2025-03-26 (initialize, no session). Each answer is one JSON object. The same answers are plain links, for an assistant without the connector: the API, with the prompt to paste into Claude or ChatGPT, starting at /api/sites. Every public fact is in the open data download.

curl -s https://rarecontext.com/mcp \
  -H 'Content-Type: application/json' -H 'Accept: application/json, text/event-stream' \
  -H 'MCP-Protocol-Version: 2025-06-18' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"search","arguments":{"query":"colossus"}}}'

Want what the pages hold back?Owners of land, grid capacity at a substation, a listed figure or an introduction to an owner: ask, and a person replies.

Request research