n8n AI Agent Workflows 2026: Build a Sales or Support Automation in an Afternoon (198k Stars)

Your AI needs to act - send emails, update CRMs, post to Slack - not just chat. n8n wires LLMs into your real business systems without code.

## n8n: Workflow Automation With Native AI Agents n8n (198,957 GitHub stars, fair-code license - free for self-hosting) is the workflow automation platform that has become the default glue for AI agents in 2026. It has 400+ integrations (Gmail, Slack, Notion, CRMs, databases, HTTP) plus native AI nodes - so an agent can read an email, decide a response with an LLM, update the CRM, and notify Slack, all in one visual workflow. ## The AI Agent Node n8n's AI Agent node is a small loop: it takes a prompt + tools, calls an LLM (OpenAI, Anthropic, Gemini, or local via Ollama), and lets the model decide which tools to call - each integration becomes a tool the agent can invoke. This turns n8n from a trigger-and-transform engine into a decision engine. A lead-qualification agent can search the CRM, check pricing, draft an email, and log the interaction - the classic demo that sells teams on it. ## A Realistic First Workflow 1. **Trigger:** New form submission (or new Gmail message). 2. **Process:** AI Agent node with instructions and access to a CRM tool + email tool. 3. **Branch:** If the LLM classifies the lead as hot, notify Slack and create a task; otherwise file it. 4. **Log:** Write the outcome to a Google Sheet for review. That is an afternoon of work and it replaces a process that previously involved three people and a shared inbox. ## Self-Hosting Reality n8n runs in Docker on a small VPS (2GB RAM is enough for light use). The free community edition is full-featured for single-team use; paid plans add SSO, LDAP, and priority support. For AI-heavy workflows, budget for the LLM calls - n8n itself is cheap, the tokens are the recurring cost. ## FAQ **Is n8n free?** Yes for self-hosting (fair-code); cloud has free and paid tiers. **Do I need to code?** No - workflows are visual; code nodes exist for the edge cases. **Can it use local LLMs?** Yes - Ollama and any OpenAI-compatible endpoint work as the agent's brain. **n8n vs Zapier vs Make?** n8n is the open-source, self-hostable choice with native AI agent support; Zapier/Make are hosted SaaS with their own AI features.