Feature Request: Webhook `Streaming Nodes` (SSE) to observe node outputs in real time

Feature Requests
1

As a developer, I want to build complex backend workflows (AI + classical) in n8n and integrate them into a small frontend app. The frontend should be able to call a workflow via Webhook and observe each node’s output in real time (not only AI nodes) to reflect progress and partial results in the UI.

Problem / Motivation

  • Many modern UIs (especially AI apps) need progressive, token/step-by-step updates instead of a single final response.

  • Today, a Webhook workflow typically returns only when the workflow finishes, which makes it hard to:

    • Show live progress (“Running X… ✓ Done”)
    • Stream intermediate results (LLM chunks, partial aggregations, logs)
    • Gracefully handle/cancel long-running operations from the UI

Proposed Solution

Add a new Respond Mode to the Webhook node: “Streaming Nodes”.

  • When enabled, the webhook initializes an SSE (Server-Sent Events) response:

    • Content-Type: text/event-stream
    • Cache-Control: no-cache
    • Connection: keep-alive
    • Flush after each event

  • During execution, n8n emits one SSE event per node (and optionally per run/item) with:

    • Node lifecycle events: node:start, node:progress, node:end
    • Workflow lifecycle events: workflow:start, workflow:heartbeat (optional), workflow:end, workflow:error

  • Each event carries a structured payload (see schema below) with safe, redacted node output.

Example (SSE wire format)

HTTP/1.1 200 OK
Content-Type: text/event-stream
Cache-Control: no-cache
Connection: keep-alive

event: workflow:start
data: {"workflowId":"abc123","executionId":"ex_789","startedAt":"2025-09-23T19:35:12.000Z"}

event: node:start
data: {"nodeId":"1","nodeName":"HTTP Request","runIndex":0,"startedAt":"2025-09-23T19:35:12.050Z"}

event: node:end
data: {"nodeId":"1","nodeName":"HTTP Request","runIndex":0,"finishedAt":"2025-09-23T19:35:12.320Z","status":"success","items":1,"outputPreview":{"body":{"message":"ok"}}}

event: node:start
data: {"nodeId":"2","nodeName":"LLM","runIndex":0,"startedAt":"2025-09-23T19:35:12.330Z"}

event: node:progress
data: {"nodeId":"2","nodeName":"LLM","runIndex":0,"progress":{"delta":"Hello "} }

event: node:progress
data: {"nodeId":"2","nodeName":"LLM","runIndex":0,"progress":{"delta":"world!"} }

event: node:end
data: {"nodeId":"2","nodeName":"LLM","runIndex":0,"finishedAt":"2025-09-23T19:35:13.100Z","status":"success","items":1}

event: workflow:end
data: {"executionId":"ex_789","finishedAt":"2025-09-23T19:35:13.110Z","status":"success"}

Why this belongs in n8n
Streaming node outputs unlocks first-class UX for AI and long-running workflows, enabling modern app patterns (progress bars, token streams, live logs) with minimal glue code. It will make n8n a far better fit for production UI integrations.

Thanks for considering!

If helpful, I’m happy to test a preview build and provide feedback on payload shape and performance.