Task Runner: Workflow hangs and crash when Code node uses $node[] with AI Agent containing postgresTool sub-nodes

[sayedgamal99]

Bug Description

When using task runners (internal or external), workflows hang and crash when all of the following conditions are met:

1. Task runners are enabled in any mode
   (N8N_RUNNERS_ENABLED=true)
2. A Code node tries to access previous node data using
   $node['nodeName'] syntax
3. The workflow contains an AI Agent with specific tool nodes
   (e.g., postgresTool)

The workflow executes successfully when any one of these conditions is removed.

To Reproduce

Environment Setup

• n8n version: 1.115.0
• Database: PostgreSQL
• Deployment: Self-hosted Docker with task runner sidecar container
• Task Runner Mode: External
• Task Runner Image: n8nio/runners:1.115.0

---

Docker Compose Configuration

services:
  n8n-prod:
    image: n8nio/n8n:1.115.0
    environment:
      - DB_TYPE=postgresdb
      - N8N_RUNNERS_ENABLED=true
      - N8N_RUNNERS_MODE=external
      - N8N_RUNNERS_BROKER_LISTEN_ADDRESS=0.0.0.0
      - N8N_RUNNERS_AUTH_TOKEN=<token>
      - N8N_SECURE_COOKIE=false
      - NODE_FUNCTION_ALLOW_BUILTIN=*
      - NODE_FUNCTION_ALLOW_EXTERNAL=*
      - N8N_BLOCK_ENV_ACCESS_IN_NODE=false
      - N8N_ENFORCE_SETTINGS_FILE_PERMISSIONS=true
      - N8N_COMMUNITY_PACKAGES_ALLOW_TOOL_USAGE=true
      # ... other configs

  task-runners:
    image: n8nio/runners:1.115.0
    environment:
      - N8N_RUNNERS_TASK_BROKER_URI=http://n8n-prod:5679
      - N8N_RUNNERS_AUTH_TOKEN=<token>
      - N8N_RUNNERS_AUTO_SHUTDOWN_TIMEOUT=0

---

Workflow Structure

Chat Trigger
  → Chat Configs (Set node)
  → Validate Request (Code node) [works - no $node[] access]
  → Input Preparation (Code node) [HANGS HERE]
  → ... other nodes
  → AI Agent [with postgresTool in tools]
  → Response

---

Code That Triggers the Issue

In the "Input Preparation" Code node:

// This causes the workflow to hang
const chatConfig = $node['Chat Configs'].json;
// ... rest of code

---

Steps to Reproduce

1. Enable task runners.

2. Create a workflow with:

   • A Chat Trigger node
   • A Set/Code node that produces configuration data
   • A downstream Code node that references the previous node using $node['nodeName']
   • An AI Agent node with tool sub-nodes (e.g., postgresTool)

3. Execute the workflow.

4. Observe that execution hangs at the Code node with $node[] reference.

and here's a sample workflow that contain the actual issue:

workflow.json

Expected behavior

Expected Behavior

The Code node should successfully access previous node data using $node['nodeName'] syntax regardless of what other node types exist in the workflow.

---

Actual Behavior

• The workflow execution hangs at the Code node.

• The following error appears in logs:

  Unrecognized node type: n8n-nodes-base.postgresTool
  

• The task runner eventually times out or crashes the workflow execution.

---

Log Output

2025-10-14 10:55:19 Registered runner "JS Task Runner" (e36iBilJnKh_Mgnvh70pV)
2025-10-14 10:55:20 Unrecognized node type: n8n-nodes-base.postgresTool

Debug Info

Debug info

core

• n8nVersion: 1.115.0
• platform: docker (self-hosted)
• nodeJsVersion: 22.19.0
• nodeEnv: production
• database: postgres
• executionMode: regular
• concurrency: -1
• license: enterprise (production)
• consumerId: c3ac9ab3-afe8-4c83-8042-895ec5a6f980

storage

• success: all
• error: all
• progress: false
• manual: true
• binaryMode: memory

pruning

• enabled: true
• maxAge: 336 hours
• maxCount: 10000 executions

client

• userAgent: mozilla/5.0 (windows nt 10.0; win64; x64) applewebkit/537.36 (khtml, like gecko) chrome/141.0.0.0 safari/537.36
• isTouchDevice: false

security

• secureCookie: false

Generated at: 2025-10-14T09:17:10.553Z

Operating System

OS: Linux (Docker)

n8n Version

1.115.0

Node.js Version

v22.19.0

Database

PostgreSQL

Execution mode

main (default)

Hosting

self hosted

[Joffcom]

Hey @sayedgamal99,

Thank you for reaching out! We’ve received your issue and are looking into it. To help us track this internally, we’ve created a Linear ticket with the reference: "GHC-5016".

We’ll keep you updated as we make progress, but if you have any additional details or context to share, feel free to add them here - it’s always helpful!

Thanks for your patience and for bringing this to our attention.

---

For INTERNAL use only: GHC-5016

[arc64]

Hi @sayedgamal99,

Can you please upgrade to 1.116.0 and let us know if you are still experiencing the issue?

[sayedgamal99]

Hi @arc64,

Unfortunately, I still have the same issue. I’ve upgraded to v1.116.0, but it’s still not working — the same conflict persists.

Here’s the error message:

2025-10-23 15:13:23 2025-10-23T12:13:23.845Z | error | Unrecognized node type: n8n-nodes-base.postgresTool {"file":"error-reporter.js","function":"defaultReport"}

[unbalanced]

The same happens to me as well on a different tool node (Unrecognized node type: n8n-nodes-base.slackTool)
On a simpler workflow that tool node simply work, and on a more complex workflow like @sayedgamal99 described, it does not work.

[equal-ericojonck]

i was having this problem today. v2.4.4
When i deleted the Supabase tool from my AI Agent, it fixed...

[Hende88]

same with Version 2.4.8, Version 2.6.2 next too, with perplexity tool.

[leoslg]

i faced the same issue, in stable version, but, using supabase tool node, which is a ai agent tool, when my first code node is execute, get stuck on the loop, and the logs contains "Unrecognized node type: n8n-nodes-base.supabaseTool"