Node Types

Nodes represent the building blocks of agentic systems. Each node has a type that determines its visual appearance and semantic meaning.

Core types

Types for agentic system components:

TypeDescriptionCommon sub_type values
agentAI agent or orchestrator—
toolExternal tool / integrationmcp, custom, api, function
modelLLM or embedding modelllm, embedding, vision, custom
memoryPersistent memory storevector, sql, file
knowledge_baseKnowledge / RAG sourcedocuments, embeddings, web_search, structured_data
promptPrompt definitionsystem, user, few_shot, template
guardrailSafety filter / policyoutput_filter, input_filter, policy
humanHuman actor / approval point—
triggerEvent trigger / webhook—
routerMessage / request router—
gatewayAPI gateway / message broker—
skillAgent skill / capability—
aggregatorData aggregator / collector—
observabilityMonitoring / logging / tracing—
channelCommunication channelweb, email, mobile, chat, app

Structural types

Types for system structure and grouping:

TypeDescriptionCommon sub_type values
systemSystem with drill-down—
componentSoftware component—
backendBackend service / datastore—
groupVisual grouping containerzone, team, vpc, boundary, custom
userUser / customer entity—
outputSystem output / response—
environmentDeployment environment—
genericUntyped node—

Visual types

Typically generated by the editor, not hand-written:

TypeDescription
noteFreeform comment
text_labelSimple text label
blockVisual block / container
imageImage display
dividerVisual divider line
waypointInvisible routing point

Example

nodes:
  orchestrator:
    type: agent
    label: Kitchen Manager
    is_looping: true

  pantry-server:
    type: tool
    sub_type: mcp
    label: Pantry MCP Server

  llm:
    type: model
    sub_type: llm
    label: Claude Opus
    provider: Anthropic

  recipe-kb:
    type: knowledge_base
    sub_type: documents
    label: Recipe Collection

  input-guard:
    type: guardrail
    sub_type: input_filter
    label: Input Validator