⚒️ Create an autonomous task-handling AI agent with OpenAI and Slack
⚡ 29 views · ⚒️ Engineering
Description
This workflow creates a self-improving AI agent inside n8n that can understand natural language tasks, plan steps, use tools (HTTP, code, search, …), reflect on results, and continue until the goal is reached — then deliver the final answer.
How it works
- Webhook or manual trigger receives a task description
- LLM creates initial plan + first tool call (or finishes immediately)
- Loop:
• Execute chosen tool
• Send observation back to LLM
• LLM reflects → decides next action or finish - When finished → format final answer, save result, send Slack notification
Setup steps
- Connect OpenAI (or Anthropic/Groq/Gemini) credential
- (Optional) Connect Slack credential for notifications
- Replace the placeholder “Other Tools” Code node with real tool nodes (Switch + HTTP Request, Google Sheets, Code node, etc.)
- Test with simple tasks first:
• “What is the current weather in Ahmedabad?”
• “Calculate 17×42 and explain the steps” - Adjust max iterations (via SplitInBatches or custom counter) to prevent infinite loops
- Activate the workflow and send POST request to webhook with JSON:
{"task": "your task here"}
Requirements
- LLM API access (gpt-4o-mini works well for testing)
- Optional: Slack workspace for alerts
Customization tips
- Upgrade to stronger reasoning models (o1-preview, Claude 3.5/3.7 Sonnet, Gemini 2.0)
- Add real tools: browser automation, vector DB lookup, file read/write, calendar
- Improve memory: append full history or use external vector store
- Add cost/safety guardrails (max iterations, forbidden actions)
Contact Us
If you need help setting up this workflow, want custom modifications, or have questions about integrating specific tools/services: 🌐 Website: https://www.oneclickitsolution.com/contact-us/
🔗 Nodes Used
HTTP Request, Slack, Webhook, OpenAI
📥 Import
Download workflow.json and import into n8n:
Workflow menu → Import from File