The Missing Layer in Odoo Automation
Most Odoo automations stop at the obvious stuff. Create a record, update a field, send an email when a status changes. Useful, but not exactly intelligent.
What if your automation could actually think? Read an unstructured email, figure out what the person is asking, query your Odoo data to find the answer, and reply — all without you touching it?
That's what this demo is about.
What is n8n?
If you haven't come across it yet, n8n is an open-source workflow automation tool. Think Zapier, but self-hostable, more powerful, and with a native AI agent node built in. It has a huge library of integrations and — crucially for this — first-class support for MCP (Model Context Protocol).
It also has a built-in Odoo node out of the box.
The Built-in Odoo Node
Before getting into the AI stuff, it's worth knowing n8n already has a native Odoo integration. You connect it with your Odoo URL, database name, and an API key from your user settings. From there you can do basic CRUD operations — create records, read data, update fields.
Check it out here: https://n8n.io/integrations/odoo/
It's solid for straightforward structured automations. But the moment your input is messy, ambiguous, or requires any kind of reasoning — it hits a wall. That's where the AI agent with MCP comes in.
The Demo: Gmail → Claude Agent → Auto-Reply
See a demo of this workflow here: https://www.youtube.com/watch?v=VxpsgyYHq7w
The workflow I built does this:
- Gmail trigger fires when a new email arrives
- The email body gets passed to a Claude AI agent node
- The Claude node connects to an MCP endpoint — my Odoo instance via OAuth
- The agent makes tool calls to Odoo to retrieve whatever data it needs
- A reply gets sent back to the original email automatically
In the live test, I sent an email asking two questions: which app had the highest view count, and how many contacts were in the system. The agent made the appropriate tool calls to Odoo, retrieved the real data, and fired back an accurate reply. No hardcoded logic. No predefined fields. It just figured it out.
Why MCP Makes the Difference
The standard Odoo node knows how to talk to Odoo, but it doesn't know what to ask or why. You have to tell it exactly which model, which fields, which filters.
With MCP, the AI agent has a set of tools it can reason about and choose from dynamically. It reads the incoming email, decides what it needs to know, picks the right tool, queries Odoo, and builds its response. The intelligence is in the agent — MCP just gives it the hands to reach into your Odoo data.
It's a fundamentally different way of building automations. Less like configuring a pipeline, more like giving an assistant access to your systems and letting it work.
What Else Could You Build?
This Gmail demo is one pattern, but the same setup works across a lot of scenarios:
- Support emails that look up order history before replying
- Incoming purchase requests that check stock and create POs
- Lead triage that scores and assigns based on CRM context
- Scheduled digests that summarise Odoo data in plain English
Anywhere you'd normally need a human to look something up before responding — that's the sweet spot.
Try It
The Odoo MCP Studio app that powers the MCP connection is available on the Odoo App Store for Odoo 19:
https://apps.odoo.com/apps/modules/19.0/odoo_remote_mcp
For n8n's built-in Odoo integration: