MCP Server: Connect Claude, VS Code, and Your AI Clients to Ontologie

What if your favorite AI tools could query, modify, and control your ontology directly? With the Ontologie MCP server (Model Context Protocol), this is now possible. Connect Claude Desktop, VS Code, Cursor, or any MCP-compatible client to your workspace and interact with your data using natural language.

Why MCP

The Model Context Protocol (MCP) is an open standard that allows AI assistants to interact with external data sources in a structured and secure way. Instead of copying and pasting data between your ontology and your AI assistant, MCP creates a direct bridge.

The use cases are numerous:

• Data exploration: ask Claude "What are the Customer entities that have no relation to a Contract?" and get a real-time answer from your ontology.
• Assisted editing: "Add a 'last_review_date' property to all Process entities" — the assistant executes the command via MCP.
• Analysis and reporting: "Summarize the status of my data validation workflow" — the assistant retrieves execution statuses and provides a synthesis.
• Development: in VS Code or Cursor, query your ontology while you code to generate code that is consistent with your data model.

What's new

157 tools across 11 categories

The Ontologie MCP server exposes 157 tools organized into 11 categories covering the entire platform:

Category	Example tools	Count
Ontology	List entities, create relations, modify properties	20+
Workflows	Execute a workflow, view results, manage schedules	15+
Agents	Invoke an agent, view history, configure tools	15+
Knowledge	Search documents, query the knowledge graph	10+
Live Data	View sources, check synchronization statuses	10+
Calendar	Create events, list upcoming deadlines	8+
Dashboard	View KPIs, retrieve statistics	8+
Spreadsheets	Read and modify spreadsheets, export data	10+
Governance	View approvals, classifications, quality status	10+
Administration	Manage members, permissions, settings	10+
Bundles	Batch operations across multiple tools	5+

Each tool is documented with its input schema, output schema, and usage examples.

Connecting to Claude Desktop

To connect Claude Desktop to your Ontologie workspace:

1. Generate an MCP API key in Settings > API Keys with the mcp.read and mcp.write scopes.
2. Configure Claude Desktop by adding the MCP server in the settings:

{
  "mcpServers": {
    "ontologie": {
      "url": "https://api.ontologie-growthsystemes.com/mcp",
      "headers": {
        "Authorization": "Bearer ont_live_votre_cle_api",
        "x-workspace-id": "votre-workspace-id"
      }
    }
  }
}

3. Test it by asking Claude a question: "List the 10 most recently created entities in my ontology."

Claude will automatically have access to all 157 tools and will use them based on your questions.

Connecting to VS Code and Cursor

MCP-compatible code editors (VS Code with the Claude extension, Cursor) are configured the same way:

1. Add the server in your editor's MCP configuration.
2. Use the tools in the assistant chat: "Show me the schema of the 'Order' entity in my ontology."
3. Generate code that is consistent with your data model: "Generate a React component that displays the properties of a Customer based on my schema."

Security and permissions

The MCP server respects the Ontologie permission system:

• API scopes: the available tools depend on your API key's scopes (mcp.read, mcp.write, mcp.workflow.execute, mcp.agent.execute).
• Workspace isolation: each connection is isolated within its workspace. No data can be accessed outside the API key's workspace.
• Audit: all actions performed via MCP are logged in the audit trail.
• SSE streaming: long-running operations (workflow execution, agent invocation) use streaming for real-time feedback.

Resources and Prompts

In addition to tools, the MCP server exposes:

• 7 Resources: direct access to structured data (ontology schema, workflow list, etc.) that AI clients can consult.
• 7 Prompts: pre-configured query templates for common use cases (ontology exploration, workflow analysis, etc.).

Getting started

1. Create an MCP API key in Settings > API Keys.
2. Select the scopes: mcp.read for read access, mcp.write for modifications.
3. Copy the configuration provided on the API key page.
4. Paste it into your MCP client (Claude Desktop, VS Code, Cursor).
5. Test it with a simple request: "How many entities does my ontology contain?"

Next steps

• MCP Server — MCP server overview
• Configuration — Connection and configuration guide
• Available tools — Full list of all 157 tools
• API Keys — Manage your keys and scopes
• Support — Need help setting up MCP?