Skip to main content

Example Workflows

Common patterns for exploring the knowledge graph through the MCP tools.

Explore a topic​

Start with a search, then drill into the most relevant node:

  1. Search for the topic:

    search_graph(query="quantum entanglement", node_type="concept")

  2. Load the node to see its definition and counts:

    get_node(node_id="<node-uuid>")

  3. Read dimensions to see how different AI models analyze the topic:

    get_dimensions(node_id="<node-uuid>")

  4. Explore connections to find related concepts:

    get_edges(node_id="<node-uuid>", limit=20)

Deep-dive into evidence​

When you want to understand the factual basis for a node:

  1. Get the facts grouped by source:

    get_facts(node_id="<node-uuid>", limit=100)

  2. Check provenance — see all sources that contributed:

    get_fact_sources(node_id="<node-uuid>")

  3. Filter by type to focus on specific evidence:

    get_facts(node_id="<node-uuid>", fact_type="measurement")

Find what a source says about a topic​

Use node intersection to find facts shared between a topic and a source entity:

  1. Search for both nodes:

    search_graph(query="climate change")
    search_graph(query="NASA", node_type="entity")

  2. Get intersecting facts:

    get_facts(
      node_id="<climate-change-uuid>",
      source_node_id="<nasa-uuid>"
    )

This returns only facts that are linked to both nodes — effectively answering "What does NASA say about climate change?"

Find connections between concepts​

Discover how two topics are related through the graph:

  1. Search for both nodes:

    search_graph(query="sleep deprivation")
    search_graph(query="immune system")

  2. Find paths between them:

    get_node_paths(
      source_node_id="<sleep-uuid>",
      target_node_id="<immune-uuid>",
      max_depth=4
    )

  3. Explore intermediate nodes on the path:

    get_node(node_id="<intermediate-node-uuid>")
    get_facts(node_id="<intermediate-node-uuid>")

Compare model perspectives​

See where AI models agree and disagree on a topic:

  1. Load all dimensions:

    get_dimensions(node_id="<node-uuid>", limit=50)

  2. Compare confidence scores — high confidence across models suggests strong evidence. Low or divergent confidence suggests uncertainty.

  3. Look for perspective nodes with different stances:

    search_graph(query="<topic>", node_type="perspective")

  4. Check edges for dialectic pairs (contradicting perspectives):

    get_edges(node_id="<perspective-uuid>")