MCP Resources - Ontologie Documentation

Overview

MCP resources are contextual data that your AI client can read to enrich its responses. Unlike tools (which execute actions), resources provide passive read-only access.

Prerequisites

API key with the MCP preset
MCP client configured (see Configuration)

Reference

The platform exposes 7 resource providers:

Ontology

URI	Description
`dataforge://ontology`	Lists available spaces
`dataforge://ontology/{espaceId}/schema`	Complete schema of a space (types, properties, relations)

Use case: Provide the ontology schema as context before a complex query.

Workflows

URI	Description
`dataforge://workflows`	Lists workflows with their status
`dataforge://workflows/{workflowId}`	Complete definition of a workflow

Knowledge Base

URI	Description
`dataforge://knowledge/sources`	Lists document sources

Agent

URI	Description
`dataforge://agents`	Lists available agents
`dataforge://agents/{agentId}`	Configuration and tools of an agent

Live Data

URI	Description
`dataforge://livedata/sources`	Lists connected sources
`dataforge://livedata/{sourceId}/schema`	Schema of a source

Calendar

URI	Description
`dataforge://agir/calendrier/calendars`	Lists calendars

Dashboard

URI	Description
`dataforge://agir/dashboards/overview`	Workspace KPI summary

Examples

Read the schema of an ontology (JSON-RPC):

{
  "jsonrpc": "2.0",
  "method": "resources/read",
  "params": {
    "uri": "dataforge://ontology/MON_ESPACE_ID/schema"
  },
  "id": 1
}

List document sources:

{
  "jsonrpc": "2.0",
  "method": "resources/read",
  "params": {
    "uri": "dataforge://knowledge/sources"
  },
  "id": 2
}

Limitations

Resources are read-only -- use tools to modify data.
Resource templates (URIs with parameters) require a valid identifier.
Resource content is filtered by workspace (multi-tenant isolation).