How to get started with an Odoo MCP server

To start with an Odoo MCP server you need four things: an Odoo database whose plan allows external API access, a dedicated Odoo user with an API key, an MCP server that talks to Odoo, and an AI client that supports MCP, such as Claude, ChatGPT, Claude Code, Cursor or VS Code. This guide covers each step, with config snippets, and was checked against official documentation on 20 October 2026.

Step 1: check prerequisites

Step 2: create the Odoo user and API key

In Odoo 20 the dialog adds a scope selector, and Odoo's own MCP server accepts keys with the "MCP" scope.

Test the key before adding MCP

On Odoo 19, a single JSON-2 call confirms the key and the user's access:

curl -s https://mycompany.example.com/json/2/res.partner/search_read \
  -H "Authorization: bearer YOUR_API_KEY" \
  -H "X-Odoo-Database: mycompany" \
  -H "Content-Type: application/json" \
  -d '{"domain": [["is_company", "=", true]], "fields": ["name"], "limit": 3}'

On Odoo 16 to 18, use XML-RPC, where the API key replaces the password:

import xmlrpc.client
url, db, user, key = "https://mycompany.example.com", "mycompany", "anna@example.com", "YOUR_API_KEY"
uid = xmlrpc.client.ServerProxy(f"{url}/xmlrpc/2/common").authenticate(db, user, key, {})
models = xmlrpc.client.ServerProxy(f"{url}/xmlrpc/2/object")
print(models.execute_kw(db, uid, key, "res.partner", "search_read",
      [[["is_company", "=", True]]], {"fields": ["name"], "limit": 3}))

If either call returns partners, the key works. If it returns an access error, fix the user's groups first.

Step 3: choose an MCP server

Step 4: connect your AI client

UI labels below are as documented on 20 October 2026. Vendors rename menus often, so follow the linked help pages if a label differs.

Claude (web and desktop): custom connector

Per Anthropic's help center, custom connectors are available on Free, Pro, Max, Team and Enterprise plans; Free is limited to one custom connector.

1. Open Customize, then Connectors, click "+ Add", then "Add custom connector".
2. Enter a name and the remote MCP server URL, for example https://mcp.example.com/mcp.
3. Choose the sign-in mode and OAuth client option the server supports, then Add.
4. On Team and Enterprise plans an Owner adds the connector under Organization settings, Connectors, and members click Connect.

Claude connects from Anthropic's cloud, so the URL must be publicly reachable over HTTPS.

Claude Desktop: local server over stdio

Edit claude_desktop_config.json: on macOS in ~/Library/Application Support/Claude/, on Windows in %APPDATA%\Claude\. This example uses the open-source mcp-server-odoo package as documented in its README; in its standard mode it expects its companion Odoo module.

{
  "mcpServers": {
    "odoo": {
      "command": "uvx",
      "args": ["mcp-server-odoo"],
      "env": {
        "ODOO_URL": "https://mycompany.example.com",
        "ODOO_API_KEY": "YOUR_API_KEY",
        "ODOO_DB": "mycompany"
      }
    }
  }
}

Restart Claude Desktop after saving. The key sits in plain text in this file, which is fine for one developer and a reason to prefer a remote server with sign-in for a team.

Claude Code

# remote server over Streamable HTTP
claude mcp add --transport http odoo https://mcp.example.com/mcp \
  --header "Authorization: Bearer YOUR_TOKEN"

# local server over stdio
claude mcp add --env ODOO_URL=https://mycompany.example.com \
  --env ODOO_API_KEY=YOUR_API_KEY --env ODOO_DB=mycompany \
  --transport stdio odoo -- uvx mcp-server-odoo

Syntax per the Claude Code MCP docs. Run /mcp inside Claude Code to check the connection or sign in.

ChatGPT

OpenAI's current documentation adds custom MCP servers from the ChatGPT Plugins page (earlier announcements called this developer mode and connectors):

1. Open Plugins, select the plus button, then "Add custom MCP server".
2. Enter a name, a description and the server URL including the /mcp path.
3. Configure authentication and confirm.

The endpoint must be public HTTPS with Streamable HTTP. Publishing a plugin to a workspace requires a workspace admin. Which plans can use write actions has changed over time, so check OpenAI's help center for your plan.

OpenAI Responses API

For your own agent code, the remote MCP tool takes a server URL and an approval policy. Keep approval on for Odoo writes:

{
  "type": "mcp",
  "server_label": "odoo",
  "server_url": "https://mcp.example.com/mcp",
  "authorization": "YOUR_TOKEN",
  "require_approval": "always"
}

Cursor and VS Code

Cursor reads .cursor/mcp.json (top-level key mcpServers). VS Code reads .vscode/mcp.json with the top-level key servers:

{
  "servers": {
    "odoo": {
      "type": "http",
      "url": "https://mcp.example.com/mcp",
      "headers": { "Authorization": "Bearer YOUR_TOKEN" }
    }
  }
}

Step 5: test with MCP Inspector

MCP Inspector is the official debugging client. It needs Node 22.19 or newer.

npx @modelcontextprotocol/inspector

Open the URL it prints (the web UI runs on port 6274 and the printed link carries an auth token). Choose Streamable HTTP for a remote server or stdio for a local one, connect, then:

1. list tools and check that write tools are the ones you expect, with sensible annotations;
2. call a read tool, for example a partner search, and compare with Odoo;
3. call a write tool on a test database only, and check that it asks for confirmation.

Common errors and fixes

Before you open the connection to more people, read the security model.