n8n-nodes-slack-ai-streaming-agent

A custom n8n community node that streams AI agent responses to Slack in real time using the native Slack AI Apps streaming API. Supports the same sub-node connections (Model / Tools / Memory) as n8n's built-in AI Agent node.

Demo

https://github.com/user-attachments/assets/78b9b67e-76cb-427a-805f-f34c83b4b932

Features

• Real-time streaming — Tokens appear in Slack as they are generated, just like ChatGPT or Claude
• No "edited" label — Uses Slack's native streaming API, not message editing
• Sub-node connections — Connect any n8n LangChain model, tool, or memory node
• Tool step display — Tool execution steps shown as task_update chunks in the Slack UI
• Automatic fallback — Falls back to chat.postMessage if streaming fails
• Feedback buttons — Optional thumbs up/down buttons on responses

Prerequisites

• n8n v1.0+ (self-hosted)
• Slack App with Agents & AI Apps feature enabled
• Node.js 18+

Installation

Option 1: npm (recommended)

Install via the n8n GUI:

1. Settings > Community Nodes > Install a community node
2. Enter n8n-nodes-slack-ai-streaming-agent
3. Click Install

Or via CLI:

cd ~/.n8n/custom
npm install n8n-nodes-slack-ai-streaming-agent

Option 2: Build from source

git clone <repository-url> n8n-nodes-slack-ai-streaming-agent
cd n8n-nodes-slack-ai-streaming-agent
npm install
npm run build

Then link to n8n:

npm link
cd ~/.n8n/custom
npm link n8n-nodes-slack-ai-streaming-agent

Option 3: Docker (volume mount)

Add to your docker-compose.yml:

services:
  n8n:
    image: n8nio/n8n
    volumes:
      - ./n8n-nodes-slack-ai-streaming-agent/dist:/home/node/.n8n/custom/node_modules/n8n-nodes-slack-ai-streaming-agent/dist
      - ./n8n-nodes-slack-ai-streaming-agent/package.json:/home/node/.n8n/custom/node_modules/n8n-nodes-slack-ai-streaming-agent/package.json

Note: Restart n8n after installation.

Slack App Setup

1. Create and configure the app

1. Create a new app at Slack API
2. Enable Agents & AI Apps under Features
3. Add the following scopes under OAuth & Permissions:

| Scope | Purpose | |---|---| | chat:write | Send messages and stream | | assistant:write | AI Apps features (status, prompts, title) | | im:history | Read DM thread history | | channels:history | Read channel thread history | | app_mentions:read | Receive @mention events in channels |

2. Subscribe to events

Enable Event Subscriptions and set the Request URL to your n8n webhook:

https://your-n8n-domain.example.com/webhook/slack-events

Subscribe to these Bot Events:

• assistant_thread_started
• assistant_thread_context_changed
• message.im
• message.channels
• app_mention

3. App Manifest

Use the template examples/slack-app-manifest.json:

1. Replace request_url with your actual n8n webhook URL
2. Paste into Slack API > App Manifests

4. Get the Bot Token

OAuth & Permissions > Install to Workspace > Copy the Bot User OAuth Token (xoxb-...).

Usage in n8n

1. Set up credentials

Uses n8n's built-in Slack credential:

1. Credentials > New Credential > Slack API
2. Enter the xoxb-... token in the Access Token field
3. Save > Test

You can reuse an existing Slack credential if you have one.

2. Build the workflow

Import the template examples/slack-ai-agent-workflow.json into n8n:

How to import: Click ... (top right) > Import from File > Select the JSON file

Workflow structure

[Webhook]
  |
  +- Respond 200 (immediate response to Slack)
  |
  +- Route by Event (Switch)
       |
       +- (assistant_thread_started)
       |   -> [Set Suggested Prompts]
       |
       +- (app_mention / message.im)
           -> [Extract Event Data]
               -> [Slack AI Streaming Agent]
                    +-- Model: OpenAI Chat Model
                    +-- Tools: (optional)
                    +-- Memory: Window Buffer Memory

Webhook settings

| Setting | Value | |---|---| | HTTP Method | POST | | Path | /slack-events | | Response Mode | Using 'Respond to Webhook' Node |

Important: Slack Events API requires a 200 response within 3 seconds. The Respond 200 node handles this immediately, allowing the workflow to continue processing asynchronously.

Slack AI Streaming Agent parameters

| Parameter | Example | Description | |---|---|---| | Channel ID | {{ $json.channel }} | Channel from the Slack event | | Thread TS | {{ $json.thread_ts }} | Thread timestamp | | Recipient User ID | {{ $json.user_id }} | User who sent the message | | Recipient Team ID | {{ $json.team_id }} | Workspace ID | | Prompt Source | Take from Previous Node | Auto-detects chatInput from input | | System Prompt | (optional) | Instructions for the AI model |

Options

| Option | Default | Description | |---|---|---| | Max Iterations | 10 | Maximum tool call iterations | | Stream Buffer Size | 64 | Characters to buffer before sending to Slack (smaller = more frequent updates) | | Feedback Buttons | false | Show thumbs up/down buttons after the response | | Set Thread Title | false | Auto-set thread title from the user's message |

3. Connect sub-nodes

Model (required)

Connect any n8n LLM model node:

• OpenAI Chat Model
• Anthropic Chat Model
• Google Gemini Chat Model

Tools (optional)

• HTTP Request Tool
• Code Tool
• Calculator
• Any custom tool

Memory (optional)

• Window Buffer Memory (in-memory, lost on restart)
• Postgres Chat Memory (persistent, recommended for production)
• Redis Chat Memory (persistent)

4. Output data

The node outputs:

{
  "channel": "D324567865",
  "thread_ts": "1724264400.000000",
  "response_text": "The full AI response text",
  "intermediate_steps": [
    {
      "toolName": "http_request",
      "toolCallId": "call_abc123",
      "args": { "url": "https://api.example.com/data" },
      "result": "..."
    }
  ],
  "token_count": 342,
  "duration_ms": 4520
}

Development

npm install       # Install dependencies
npm run build     # Build
npm run lint      # Type check
npm test          # Run tests
npm run dev       # Watch mode

Limitations

• Streaming only works within threads (DMs are always threaded)
• Block Kit cannot be used during streaming (only on stopStream)
• Unfurls are disabled during streaming
• Workspace guests cannot access the Agents & AI Apps feature
• Slack AI Apps requires a paid plan (Developer Program offers a free sandbox)

License

MIT