Businesses are increasingly looking for ways to connect AI assistants directly to their ERP systems.
Instead of exporting Odoo data into spreadsheets and then asking an AI tool to analyse it, Model Context Protocol (MCP) provides a way for compatible AI clients to interact with external tools and business data.
With Odoo 20, the Odoo database can act as an MCP server. An external AI client can connect to it, discover available tools, send requests, and receive results. Depending on the tools and permissions configured, the AI can read information or perform selected actions in Odoo.
This opens the door to use cases such as:
AI-powered Odoo reporting
Natural-language ERP search
Sales pipeline analysis
Inventory analysis
Customer data lookup
Business reporting
AI workflow automation
AI agents for ERP
Controlled record creation and updates
But how do you actually connect an AI assistant such as Claude or ChatGPT to Odoo using MCP?
Let's walk through the architecture, requirements, setup process, security considerations and practical business use cases.
What Is MCP?
Model Context Protocol (MCP) is an open standard that allows AI applications to communicate with external systems through defined tools and data connections.
In an Odoo environment, MCP creates a structured communication layer between an AI client and your ERP.
A simplified architecture looks like this:
User → AI Client → MCP Client → Odoo MCP Server → Odoo Database
For example, instead of manually opening CRM and filtering opportunities, a sales manager could ask:
“Show me my open opportunities above £10,000 and group them by sales stage.”
The AI client can send that request to the Odoo MCP server, use an available tool, retrieve the information and return the result in natural language.
Odoo's documentation describes this workflow as an external AI agent connecting to the Odoo database, discovering available tools, invoking the appropriate tool and receiving the result.
Why Connect AI to Odoo?
Traditional ERP workflows require users to navigate menus, filters, reports and records.
MCP creates another way to interact with Odoo.
Instead of:
Open Odoo → Find Module → Apply Filters → Search Records → Analyse Data
the interaction can become:
Ask AI → AI Uses Odoo Tool → Receive Result
This can be particularly useful for businesses that already have large amounts of operational information stored in Odoo.
Potential applications include:
Sales
Review open opportunities
Analyse sales pipelines
Search customer information
Compare sales performance
Inventory
Check stock availability
Review inventory information
Identify products requiring attention
Analyse demand information
Finance
Review invoice information
Compare historical sales
Analyse financial records
Search accounting information
Operations
Retrieve business records
Search workflows
Create approved records
Update selected information
Odoo's current MCP documentation provides examples covering sales pipeline analysis, manufacturing and inventory planning, historical reporting, bulk record creation, filtering and data lookup/editing.
What You Need Before Connecting Odoo to MCP
Before starting, make sure your environment meets the required technical conditions.
For the native Odoo 20 MCP workflow, you generally need:
An Odoo 20 database
An MCP-compatible AI client
An Odoo user with appropriate permissions
An Odoo API key with MCP scope
The Odoo MCP endpoint
A compatible MCP client configuration
Node.js when using the npx approach documented by Odoo
Odoo's current documentation states that the MCP setup requires an MCP-compatible AI agent and Node.js for running the npx command used in its connection examples.
The exact requirements can vary depending on whether you are using Claude, ChatGPT, another AI client, or a custom MCP application.
Step 1: Create an Odoo MCP API Key
Authentication is the first important step.
In Odoo, open your user preferences and go to the security section.
Create a new API key and select the appropriate MCP scope.
Odoo's documentation explains that the API key authenticates the MCP client and is used to identify the user's permissions when requests are sent to the Odoo database. The generated key is shown only once, so it needs to be stored securely.
The general process is:
Odoo → My Preferences → Security → Add API Key → MCP Scope → Create Key
Do not paste your API key into a public document, website, GitHub repository or unsecured configuration file.
Step 2: Find Your Odoo MCP Endpoint
The Odoo MCP endpoint follows your database URL.
For example:
https://your-company.odoo.com/mcp
The /mcp endpoint is the address that the compatible AI client uses to communicate with the Odoo MCP server.
Odoo's documentation specifically describes creating the endpoint by adding /mcp to the database URL.
Your architecture now looks like:
Claude / ChatGPT → HTTPS MCP Endpoint → Odoo → Business Data
Step 3: Connect Claude to Odoo MCP
Claude is one of the AI clients specifically documented by Odoo for MCP connections.
Odoo currently provides instructions for connecting through Claude Desktop and Claude Code CLI.
Claude Desktop
In Claude Desktop, open the developer settings and locate the local MCP server configuration.
Odoo's documented configuration uses npx and mcp-remote to connect Claude to the Odoo MCP endpoint.
The architecture is:
Claude Desktop → mcp-remote → Odoo /mcp endpoint
The configuration includes:
Your Odoo MCP endpoint
The authentication header
Your Odoo API key
A custom MCP server name
After saving the configuration, Claude should show the MCP server and its connection status.
Claude Code
If you use Claude Code, Odoo provides a CLI-based approach.
The connection follows this general structure:
Claude Code → MCP configuration → Odoo MCP Server
After adding the server, you can use the MCP status command to verify the connection.
Odoo's documentation provides the current CLI syntax and configuration approach.
This distinction is important because Claude Desktop and Claude Code are not configured in exactly the same way.
Step 4: Connect ChatGPT to Odoo MCP
ChatGPT requires a slightly different approach.
OpenAI currently supports remote MCP servers and custom MCP apps, but availability depends on the ChatGPT plan and workspace configuration. OpenAI's current documentation states that full MCP support, including write/modify actions, is rolling out for Business, Enterprise and Edu plans, while Pro users can use MCP connections with read/fetch permissions in developer mode.
For a business using ChatGPT with a custom Odoo MCP application, the general workflow is:
ChatGPT → Developer Mode → Custom MCP App → Odoo MCP Server
The MCP server must be reachable through an appropriate remote HTTPS connection.
OpenAI's current MCP documentation also distinguishes remote MCP servers from local/private MCP servers and provides a Secure MCP Tunnel option for environments where the MCP server should not be exposed directly to the public internet.
This is important for businesses running Odoo on-premise or inside a private network.
ChatGPT Connection Considerations
Before connecting ChatGPT to Odoo, check:
Your ChatGPT plan
Workspace administrator permissions
Developer mode availability
Whether your MCP server is remotely reachable
Whether you need read-only or write access
Whether your organisation permits custom MCP applications
Do not assume that a ChatGPT account automatically supports every MCP capability.
Step 5: Expose Odoo Tools to the AI Client
Connecting the MCP server is only part of the process.
You also need to decide what the AI is allowed to do.
Odoo exposes several tools by default, including tools for retrieving fields, models, initial context, searching and reading groups.
Additional tools can be manually exposed through Odoo's server actions configuration.
Odoo's documentation explains that tools can be marked as available through MCP and can optionally be marked as Readonly Tools.
This creates an important distinction:
Connected does not mean unrestricted.
You might allow an AI agent to:
Search customers
Read sales orders
Analyse inventory
while preventing it from:
Deleting records
Modifying financial information
Creating purchase orders
Changing sensitive customer information
unless those actions have been deliberately configured.
Step 6: Test the Connection With Read-Only Queries
Don't begin by allowing AI to modify your ERP.
Start with simple read-only requests.
For example:
“Show me my open sales opportunities.”
Then:
“Group the opportunities by sales stage.”
Then:
“Which opportunities have the highest expected revenue?”
Then:
“Compare this month's sales with last month's sales.”
These tests help verify:
Authentication
Tool discovery
Odoo permissions
Data retrieval
AI interpretation
MCP connectivity
Odoo itself provides sales pipeline, inventory planning and historical reporting as examples of MCP use cases.
Step 7: Move From Search to Automation
Once read-only access works correctly, you can consider more advanced workflows.
For example:
Customer Search
→ Find customer
→ Check recent orders
→ Review payment information
→ Prepare follow-up
Another workflow could be:
Sales Analysis
→ Retrieve opportunities
→ Group by salesperson
→ Identify high-value opportunities
→ Create follow-up activities
The final step should only be enabled if the appropriate Odoo tools, permissions and approval controls are in place.
This is where MCP becomes particularly interesting for AI-powered ERP automation.
Practical Odoo + AI MCP Use Cases
Sales Pipeline Analysis
Ask:
“Show me open opportunities by salesperson and highlight opportunities that have not been updated recently.”
This can reduce manual CRM reporting.
Inventory Planning
Ask:
“Which products have low forecasted availability this month?”
Odoo's MCP documentation specifically provides manufacturing and inventory planning as an example use case.
Historical Reporting
Ask:
“Compare our invoiced revenue from this quarter with the previous quarter.”
This can make business reporting more conversational.
Customer Data Lookup
Ask:
“Check whether this customer has a tax ID in Odoo.”
With the appropriate tool and permissions, the AI can retrieve the relevant information.
Controlled Record Creation
Once write-capable tools are deliberately exposed, an AI workflow can potentially assist with creating records.
For example:
“Create contacts for these five leads.”
Odoo lists bulk record creation among its MCP examples.
However, write actions should be introduced carefully.
Odoo MCP Security: What Businesses Should Consider
Connecting AI to an ERP creates a new access layer to business information.
That means security should be designed before automation.
Consider:
1. User Permissions
The AI should operate within the appropriate user's Odoo permissions.
2. API Key Management
API keys should be stored securely and rotated according to your organisation's security policy.
3. Read vs Write Access
Start with read-only access wherever possible.
4. Tool Exposure
Only expose the Odoo tools that are genuinely required.
5. Approval Workflows
Sensitive actions should require human approval where appropriate.
6. Auditability
Businesses should be able to understand which actions were performed and by whom.
Odoo's documentation specifically notes that marking a tool as read-only tells the AI client that it is safe to invoke without explicit user approval; it does not hide the tool from the client.
This is why Odoo MCP security should be treated as part of the implementation rather than an afterthought.
Odoo MCP vs Traditional Odoo API Integration
MCP does not replace every type of Odoo integration.
Traditional API integrations remain useful when two systems need structured, predictable and repeatable data exchange.
For example:
Shopify → Odoo → Inventory → Delivery
is a traditional integration workflow.
MCP is more useful when an AI needs to interact with tools and business data dynamically.
For example:
User → AI → Odoo Data → Analysis → Suggested Action
Both approaches can exist together.
ERP Artists provides Odoo Integration Services for connecting Odoo with eCommerce, payment, CRM, shipping, productivity and AI platforms.
When Should You Connect AI to Odoo?
MCP is worth considering when your employees repeatedly spend time:
Searching ERP records
Preparing reports
Checking customer information
Analysing sales data
Reviewing inventory
Processing documents
Performing repetitive ERP tasks
Moving information between systems
But don't start with:
“How can we connect AI to everything?”
Start with:
“Which business process wastes the most time?”
Then determine whether MCP and AI can improve that process safely.
How ERP Artists Can Help With Odoo AI and MCP
Connecting an AI client to Odoo is only the technical beginning.
The bigger challenge is deciding:
Which business processes should be automated?
Which Odoo models should AI access?
Which tools should be exposed?
Which actions should remain read-only?
Where should human approval be required?
How should external systems connect?
How should the solution be maintained?
ERP Artists works across AI for ERP, Odoo development, Odoo integration, and Odoo implementation.
For businesses exploring MCP, the practical approach is to start with a controlled workflow, test it using real business requirements, establish permissions and governance, and then expand the AI capabilities gradually.
Final Thoughts
Odoo MCP creates a bridge between AI assistants and ERP data.
With the right configuration, compatible AI clients such as Claude and supported ChatGPT MCP applications can interact with Odoo through defined tools rather than relying on manual data exports.
The potential applications are significant:
AI Search → Reporting → Analysis → Workflow Assistance → Controlled Automation
But MCP should not be treated as a shortcut around ERP security.
The most effective implementation starts with a specific business problem, uses the minimum required permissions, exposes only necessary tools and introduces write actions carefully.
For businesses already using Odoo, this can turn AI from a separate productivity tool into a more connected part of the ERP environment.
