MCP OKN Server

A Model Context Protocol (MCP) server providing seamless access to SPARQL endpoints with specialized support for the OKN knowledge graphs. Query 30+ scientific knowledge graphs using natural language through your favorite AI assistant.

⚠️ Endpoint updated — action required

The hosted MCP endpoint has moved to https://apps.okn.us/okn-mcp-dev/mcp.

If you previously connected the mcp-proto-okn server (the connector named proto-okn, pointing at the old https://apps.okn.us/okn-mcp/mcp endpoint), delete or deactivate that connector in your AI agent and add a new connector pointing at the new URL above. The setup steps below use the new connector name mcp-okn-dev.

Transition period: the legacy mcp-proto-okn server (old endpoint) will remain available for a transition period, so any existing setups will keep working until you have migrated. We recommend moving to the new endpoint as soon as possible to take advantage of the cross-graph query capabilities.

What it does

The MCP OKN Server connects AI assistants — such as Claude Desktop, ChatGPT, VS Code Copilot, and others — directly to the FRINK-hosted knowledge graphs. Instead of writing SPARQL by hand, you describe what you need in plain English and the server translates your request into structured queries across one or more graphs.

Key features

FRINK integration — automatic detection and documentation linking for all FRINK-hosted knowledge graphs
Federated queries — query multiple knowledge graphs simultaneously in a single prompt
Precomputed crosswalks — verified recipes for joining two or three knowledge graphs on a shared identifier, so questions one graph can’t answer alone become answerable across graphs
Worked examples and transcripts — a catalog of example questions for each crosswalk, every one backed by a replayable transcript of the prompt, answer, and verbatim SPARQL
Ontology-driven search expansion — queries automatically expand using ontology hierarchies (MONDO, HP, GO, UBERON)
Automatic documentation — registry links and metadata for every OKN knowledge graph

How to connect

A hosted instance of the assistant endpoint is available at https://apps.okn.us/okn-mcp-dev/mcp. Point your MCP-capable client at that URL — no local installation required. Visiting it directly in a browser returns a JSON-RPC error (the endpoint speaks JSON-RPC, not HTTP GET); that is expected and does not indicate the server is down. Setup steps below cover the two most common clients.

What this enables

Once connected, your assistant can discover OKN graphs, inspect graph metadata, run SPARQL queries, bridge identifiers, and return answers with source graph context. It is most useful for open-ended cross-graph questions.

Quick verification prompt

“List all OKN knowledge graphs as a table of shortname and description.”

Claude Desktop

Requirements: A Claude Pro or Max subscription is required for MCP connectors in Claude Desktop.

Download Claude Desktop from claude.ai/download.

Launch Claude Desktop and open Claude → Settings.
Choose Connectors from the Settings menu. If you previously added the legacy mcp-proto-okn server, find the proto-okn connector and delete or deactivate it.
Click Add custom connector and enter:
- Name: mcp-okn-dev
- URL: https://apps.okn.us/okn-mcp-dev/mcp
Once mcp-okn-dev appears in the connector list, click Configure and set tool permissions to Always allow.
Start a new chat, click the + button, toggle mcp-okn-dev on, and turn Web search off.
Try the verification prompt above. If the assistant returns a graph list, the connector is working.

ChatGPT

Requirements: Full MCP app support is available for ChatGPT Business, Enterprise, and Edu workspaces. ChatGPT Pro users can use custom MCP apps in developer mode. Custom MCP apps are web-only, so use the ChatGPT web app at chatgpt.com rather than the desktop app.

Note: ChatGPT may not perform as well as Claude models for this task. For best results, use Claude Sonnet or Claude Opus.

Sign in to chatgpt.com.
Click your profile name/avatar.
Open Settings.
Go to Apps.
Open Advanced settings.
Turn on Developer mode (required for custom MCP apps in ChatGPT).
Return to Apps.
Click Create app.
Enter the MCP app details:
- Name: mcp-okn-dev
- URL: https://apps.okn.us/okn-mcp-dev/mcp
Save or create the app.
Start a new chat.
Click the + button in the message box.
Select mcp-okn-dev from the list of available apps/tools.
Turn Web search off before testing, so ChatGPT uses the MCP app rather than web search.
Run the verification prompt. If ChatGPT returns a graph list, the MCP app is working.

Claude Code

Add to .mcp.json in your project root (or ~/.claude/settings.json for all projects). If you previously configured the legacy proto-okn server, remove that entry first.

{
  "mcpServers": {
    "mcp-okn-dev": {
      "type": "url",
      "url": "https://apps.okn.us/okn-mcp-dev/mcp"
    }
  }
}

Verify with /mcp — you should see mcp-okn-dev connected.

Compatible clients

Any MCP-compliant application works. Popular choices include:

Claude Desktop
Claude Code
ChatGPT
VS Code with GitHub Copilot
Cursor

Example queries

Once connected, try prompts like these:

“List all Proto-OKN knowledge graphs as a table of shortname and description.” — Result
“List all verified crosswalks, grouped by domain, with an example of what each answers.” — Result
“Give a high-level overview of the spoke-genelab knowledge graph — its main classes and relationships — and draw the schema diagram.” — Result

Review the full set of single-KG and cross-KG queries, along with their chat transcripts.

If you find an issue or would like to contribute additional example queries, please file an issue.