DEDUPED Hot AI repos, turned into use cases Get the Monday 5

How to build an AI agent workflow with a human approval step with n8n

Chain a trigger, a model step, tools and an approval so the agent drafts and a person signs off before anything goes out.

Short answer: Start from a template in the 9,000+ workflow library or a blank canvas; then add a trigger, then an AI step using OpenAI, Anthropic, Google or an open-source model; give the agent tools from your integrations. All steps.

Credentials. Each connected app stores its credentials in n8n; restrict who can open the instance.

Who it is forA team that wants AI to do the busywork on emails, tickets or leads but never act without a check.

The problemFully automatic AI flows are risky and fully manual work is slow.

The outcomeA multi-step workflow where the AI step proposes an action and waits for approval before the final step runs.

What you need first

  • Docker for the self-hosted quick start
  • Accounts or API keys for the apps and models you connect

Steps

  1. Start from a template in the 9,000+ workflow library or a blank canvas
  2. Add a trigger, then an AI step using OpenAI, Anthropic, Google or an open-source model
  3. Give the agent tools from your integrations
  4. Insert a human approval step before any step that sends or changes data
  5. Run it on test data and read the execution log before turning it on

Prompt for your agent

In my n8n-io/n8n instance, build a workflow that triggers on <event>, uses <model> to draft <output>, waits for my approval, then <final action>. Explain each node and where credentials are stored.

Caveat.

Use this in your AI agent

Ask your agent

Using the Deduped MCP, get the setup steps for n8n-io/n8n and "Build an AI agent workflow with a human approval step"

Add Deduped to Claude Code once

claude mcp add --transport http deduped https://dedupedai.com/mcp

Cursor, ChatGPT, others

Share: X LinkedIn Hacker News Email Copy link

Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer