Use OpenHaven with your AI assistant

The server address

An assistant that supports MCP (the Model Context Protocol) needs this address. The server is read-only and needs no sign-in or key.

https://openhaven.net/mcp

Set up your assistant

Pick your assistant. Each one links to its maker's own guide, which is the place to check if a menu has moved.

Claude (the app and the web)

Open Customize, then Connectors.
Choose Add, then Add custom connector.
Enter OpenHaven as the name and the server address above as the URL.
The server needs no sign-in or key, so leave those empty.

Claude's guide to custom connectors (Opens in a new tab)

Claude Code (the command line)

Run this once:

claude mcp add --transport http openhaven https://openhaven.net/mcp

Claude Code's guide to MCP servers (Opens in a new tab)

ChatGPT

Custom MCP servers are available in ChatGPT through developer mode, on some plans and workspaces. OpenAI’s guide has the current steps and where the setting is.
When ChatGPT asks for the server URL, enter the server address above.
The server needs no sign-in or key.

OpenAI's guide to custom MCP servers in ChatGPT (Opens in a new tab)

Cursor

Add this to .cursor/mcp.json in a project, or to ~/.cursor/mcp.json for every project:

{
  "mcpServers": {
    "openhaven": {
      "url": "https://openhaven.net/mcp"
    }
  }
}

Cursor's guide to MCP (Opens in a new tab)

Any other MCP client

Choose a remote server (Streamable HTTP) and enter the server address above. No sign-in or key is needed.

To run the server on your own computer instead (stdio), build it from the OpenHaven repository (see mcp/README.md) and point your client at it:

{
  "mcpServers": {
    "openhaven": {
      "command": "node",
      "args": [
        "/path/to/openhaven.net/mcp/dist/index.js",
        "--stdio"
      ]
    }
  }
}

What you can ask

Questions the server is built to answer. Every answer names the page or record it came from.

Which tools fit consent-based governance, and what does OpenHaven say about each?
What does OpenHaven say about mutual credit clearing?
Which tools could I use for proof of personhood, and how do they compare on capture risk?
Compare two Monetary Architecture mechanisms and show me the pages they come from.

What it will not do

It is read-only. Nothing can be written, submitted or changed through it.
It only repeats what the public openhaven.net site says, and each answer carries the page it came from.
Interactive views (the Navigator, the Matrix, the Graph Explorer and the Monetary Architecture and Organization & Governance Model apps) have no text beyond their data, which it reads through its data tools.
Images and diagrams are not interpreted; their alt text is kept where it exists.
Relationships come from a published snapshot of the graph, not a live database.
Requests are rate limited for each client address.

What the server contains

Its tools, by what they cover:

The whole site: page map, search, reading any page: list_pages search_site get_page list_research_reports
Tech tools (the matrix): search_tech_tools get_tech_tool
Use case domains, affordances and entity types: list_use_case_domains list_affordances list_entity_types
The Navigator path: from a use case to tools: find_tools_for_use_case
Monetary Architecture and Organization & Governance Models: list_monetary_mechanisms get_monetary_mechanism list_org_governance_models get_org_governance_model get_data_dictionary
Relationships between things, with where each one comes from: get_relationships list_relationship_types

Not using MCP?

The same data is available without an assistant as static JSON files: API for Developers and a short index for language-model agents: llms.txt.