Platform 1 of 12·OS Layer 3 (Orchestration)·Required in all deployment modes

Scope of this page. Reference material for the n8n platform: what it is, what it is normally used for, and how and why Synedrix OS uses it. It is descriptive, not normative — where this page and a canonical spec in docs/os/ disagree, the canonical spec wins. Platform responsibilities are set by Tech Stack; the build-level mapping is n8n Implementation Guide.

What n8n is

n8n describes itself as “a fair-code licensed workflow automation tool that combines AI capabilities with business process automation.”

It is a self-hostable workflow engine with a visual editor. The model is small and consistent:

ConceptOfficial definition
Workflow“A collection of nodes that automate a process. Workflows begin execution when a trigger condition occurs and execute sequentially to achieve complex tasks.”
Node“Individual components that you compose to create workflows. Nodes define when the workflow should run, allow you to fetch, send, and process data, can define flow control logic, and connect with external services.”
Trigger node“A special node responsible for executing the workflow in response to certain conditions. All production workflows need at least one trigger.”
Credential“Stores authentication information to connect with specific apps and services.”
ExpressionPopulates node parameters dynamically by executing JavaScript against data from previous nodes, other workflows, or the environment.
Cluster node“Groups of nodes that work together”: a root node plus one or more sub-nodes that extend it. This is how n8n models AI components.

The AI nodes are an implementation of LangChain’s JavaScript framework. n8n maps LangChain concepts onto cluster nodes: Agent and Chain are root nodes; language model, memory, tool, retriever, embeddings, output parser and text splitter are sub-nodes. Memory sub-nodes attach only to the AI Agent root node — n8n’s chain nodes do not support memory.

What n8n is normally used for

The general-purpose cases n8n’s own documentation leads with:

  • Connecting systems that do not talk to each other — moving and reshaping data between APIs, databases and SaaS tools, with credentials and retries handled by the engine rather than by hand-written scripts.
  • Creating API endpoints — exposing a workflow over a webhook so other systems can call it.
  • Scheduled and event-driven automation — time-based triggers and webhook triggers as the entry point to a process.
  • AI agent chat systems — building conversational agents from the AI Agent root node plus model, memory and tool sub-nodes.
  • AI-assisted data work — scraping, summarising, classifying and enriching content as steps inside a larger automated process.

The common thread is that n8n owns sequencing and integration: what happens, in what order, on what trigger, with what retry behaviour — not the business logic of any one step.

How Synedrix OS uses it

n8n is the orchestration kernel of Synedrix OS and sits at Layer 3 of the twelve-layer taxonomy. It is required in all three deployment modes.

flowchart LR
  n1([n8n]):::term
  subgraph What it owns
    n2[Agent invocation routing]:::agent
    n3[Trigger and action composition]:::agent
    n4[Retry and timeout management]:::agent
    n5[Approval routing]:::agent
    n6[Sub-workflow calls]:::agent
  end
  n1 --> n2
  subgraph What it does not own
    n7[Stateful cyclic multi-agent reasoning]:::warn
    n8[LangGraph]:::term
    n7 --> n8
    n9[Policy decisions]:::warn
    n10[Policy Guard]:::term
    n9 --> n10
    n11[Canonical business state]:::warn
    n12[Supabase PG17]:::term
    n11 --> n12
    n13[LLM traces prompts evaluations]:::warn
    n14[Langfuse]:::term
    n13 --> n14
  end
  n1 --> n7
  subgraph Why n8n
    n15[Self-hostable in full]:::term
    n16[It can be isolated]:::term
    n17[The AI nodes are LangChain underneath]:::term
    n18[Visual execution graphs are an operator surface]:::term
  end
  n1 --> n15

What it owns

Per Tech Stack, n8n owns:

  • Workflow execution and the execution graph for all domain-pack workflows
  • Agent invocation routing
  • Trigger and action composition
  • Retry and timeout management
  • Approval routing
  • Sub-workflow calls

Every workflow declared under the Workflow Contract Spec executes here. The contract’s trigger_type, retry_policy, error_policy and approval_requirements fields are n8n concerns directly.

What it does not own

This boundary is the point of ADR-OS-001, which split the Agent Runtime (Layer 4) out of Orchestration (Layer 3):

ConcernOwner
Stateful, cyclic, multi-agent reasoning (iterative, consensus, reflect-and-revise, hierarchical)LangGraph (Layer 4)
Policy decisionsPolicy Guard (os-agt-002)
Canonical business stateSupabase PG17 (Layer 7)
LLM traces, prompts, evaluationsLangfuse (Layer 10)

Simple acyclic agents may still run as n8n AI Agent nodes. Anything stateful or cyclic is invoked as a governed step against the LangGraph runtime over HTTP/MCP/A2A, carrying correlation_id, causation_id, the resolved active_tool_set and deployment_mode, and returning the canonical Outcome Envelope. The boundary contract is specified in Orchestration & Runtime Strategy §5, with a worked reference implementation in POC: n8n → LangGraph Governance Bridge.

How it is operated

n8n’s queue mode is the scaling model relevant to Synedrix OS. In queue mode a main instance handles triggers and webhook calls and generates (but does not run) an execution; the execution ID goes to Redis, which holds the pending queue; a worker picks it up, reads workflow information from the database, executes, writes results back and posts completion to Redis.

This is worth noting architecturally: n8n’s own scaling path uses Redis as a message broker, which is the same platform Synedrix OS runs at Layer 7 as its coordination bus. The two roles are distinct — n8n’s execution queue is infrastructure, the OS coordination bus is application state — and should not be collapsed onto one logical database without deliberate decision.

n8n also supports OpenTelemetry tracing via environment variables on self-hosted instances, which is the supported path for feeding execution telemetry into the Layer 10 observability platforms rather than reading it out of n8n’s own UI.

Why n8n

  • It is a workflow engine, not an agent framework. Synedrix OS needs Layer 3 to be boring, inspectable and deterministic. n8n sequences and retries; it does not reason. That is exactly the separation ADR-OS-001 formalises.
  • Self-hostable in full. The Community edition runs “with almost the complete feature set” for free, indefinitely, with no licence key — which is what Modes 2 and 3 require.
  • It can be isolated. n8n documents an explicit configuration for preventing an instance from connecting to n8n’s own servers, which is a hard requirement for air-gapped Mode 3.
  • The AI nodes are LangChain underneath. Simple agents built as n8n AI Agent nodes and complex agents built in LangGraph share a conceptual vocabulary, which keeps the Layer 3 / Layer 4 split a deployment decision rather than a rewrite.
  • Visual execution graphs are an operator surface. A governed system needs humans to be able to see what ran. The canvas and execution history are that surface for Layer 3.

Deployment and licensing

Self-hosting. Documented install paths are Docker, Docker Compose, npm, and Kubernetes, plus cloud-provider guides (AWS, Azure, Google Cloud Run, Google Kubernetes, Digital Ocean, Hetzner, OpenShift). Self-hosted editions are Community (free), Registered Community (free, unlocks folders, debug-in-editor and custom execution data by registering an email), Business and Enterprise (paid, licence key).

Licence — read this carefully. n8n is fair-code, not open source. Per n8n’s LICENSE.md:

  • Most of the codebase is under the Sustainable Use License v1.0.
  • Source files containing .ee. in the filename or .ee in the dirname are not covered by it and require a valid n8n Enterprise License.
  • The Sustainable Use License limits use as follows: “You may use or modify the software only for your own internal business purposes or for non-commercial or personal use. You may distribute the software or provide it to others only if you do so free of charge for non-commercial purposes.”

Open item. Tech Stack currently records n8n as “open source under Apache 2.0” and states that every component in the stack is open source. Neither is accurate for n8n as licensed today. The Sustainable Use License’s restriction on providing the software to others other than free of charge for non-commercial purposes is material to any deployment model where Synedrix OS is delivered to a customer as a product. This needs an explicit decision and a correction to the canonical stack document — it is not resolved by this page.

Official documentation

  • Synedrix OS Tech Stack — normative platform responsibilities; §1 covers n8n.
  • ADR-OS-001 — Agent Runtime Split — the decision that reduced n8n to pure orchestration.
  • Orchestration & Runtime Strategy — the n8n↔LangGraph boundary and governance bridge.
  • n8n Implementation Guide — agent spec to AI Agent node, outcome envelope assembly, correlation ID propagation.
  • Workflow Contract Spec — the contract every n8n workflow must satisfy.
  • Layer Taxonomy Table — Layer 3 definition and platform mapping.
  • POC: n8n → LangGraph Governance Bridge — WF-OS-POC-001, the first end-to-end Layer 3 → Layer 4 invocation.

Synedrix OS · Platform reference · n8n