More businesses are asking the same question: what if our AI assistant could actually do something with our business data, not just talk about it? Tools like Claude, ChatGPT, and Gemini have become genuinely useful for reasoning, drafting, and analysis but the moment you want them to check a real invoice, update a lead, or pull live sales figures, they hit a wall. The data lives in your ERP, and the AI has no safe way in.
That's the tension. Connecting a large language model directly to a live ERP database is not something to do casually. Without the right controls, you risk exposing sensitive records, letting an AI perform unintended actions, or bypassing the access rules your team has spent years refining.
This is exactly the gap the Odoo MCP Server is designed to close, giving AI assistants a secure, governed way to interact with Odoo, without compromising the safeguards already built into the system.
What is Odoo Remote MCP Server?
Odoo Remote MCP Server is a module built by Creyox Technologies that turns an Odoo instance into a fully-fledged MCP (Model Context Protocol) server. MCP is an open standard that allows AI assistants to connect to external systems in a structured, secure way rather than relying on brittle scripts or one-off integrations.
With this module installed, an Odoo database can expose its data and functionality to any MCP-compatible AI client. In practice, this means you can connect Claude to Odoo along with ChatGPT, Gemini, or any other MCP-aware assistant and have it read records, generate reports, or make updates using plain natural language, all while staying inside a controlled, auditable environment. At the core of this is Odoo natural language CRUD: the ability to create, read, update, and delete records simply by describing what you want done.
Key Features
- Universal MCP-client compatibility — works with any AI assistant that supports the Model Context Protocol, so you're not locked into a single vendor.
- Granular tool allowlists and IP restrictions decide exactly which actions an AI client is permitted to perform, and from which networks it can connect.
- PII and sensitive-field masking — automatically hide or obscure fields like emails, phone numbers, or other personal data before they ever reach the AI.
- Natural-language CRUD on Odoo records — create, read, update, or delete records conversationally, without writing code or navigating menus.
- Serverless React dashboard generation — ask an AI to assemble a working dashboard from live Odoo data, generated on demand.
- ECharts analytics visualizations — request charts and graphs directly from a prompt, built from real-time figures.
- Background job queue and audit logging — longer-running tasks are processed asynchronously, and every action is logged for traceability.
How It Works
At a conceptual level, every request from an AI assistant passes through a secure ingress gateway before it touches any Odoo data. The gateway inspects the requested tool call in a sandboxed context, checking it against the allowlists and IP rules configured for that connection.
If the request involves sensitive fields, masking rules are applied before any data leaves the system. From there, the action whether it's a read, a write, or a more complex operation is executed through Odoo's native ORM, which means it's still subject to the same access rights and record rules that apply to any other user or integration. Finally, the interaction is written to an audit log, so there's a clear record of what the AI did and when.
Nothing here reinvents Odoo's security model. It simply gives AI assistants a properly gated entry point into it, the kind of controlled, auditable foundation that a serious Odoo AI agent integration needs before it's trusted with real business data.
Use Cases
- A sales team asking for a pipeline summary. Instead of building a report, a sales manager can ask their AI assistant to summarize open opportunities by stage, and get an answer pulled directly from live CRM data.
- A support team checking order status conversationally. A support agent can ask whether an invoice was paid or a delivery has shipped, without switching screens or running a search.
- A developer scaffolding a custom module. Using natural language, a developer can have an AI assistant draft the skeleton of a custom Odoo module, speeding up the early stages of development.
- An analyst requesting an instant chart. Rather than building a dashboard from scratch, an analyst can ask for a chart of quarterly revenue by region and receive a live visualization built from current data.
Why Security Matters Here
For any IT decision-maker, the natural concern is: does giving an AI access to Odoo mean giving up control? With this module, the answer is no. Every action an AI assistant takes still passes through Odoo's existing access rights, record rules, and permission structure, the same ones already governing your human users.
This is what makes secure AI access to ERP data possible in the first place. The AI isn't granted a special back door; it operates within the same boundaries as any other authenticated user, further narrowed by the tool allowlists, IP restrictions, and field masking configured for that specific connection. Combined with audit logging, this gives IT teams visibility into exactly what was accessed and changed, and by which client.
Getting Started
Getting up and running follows a fairly straightforward path. After installing the module, you configure a server profile that defines which models, fields, and actions are exposed, along with the access rules allowlists, IP restrictions, and masking policies that apply to that profile. Once configured, you connect your AI client of choice using the generated credentials, and it can begin interacting with Odoo through natural language.
For the full module details or to see it in action, visit the listing on the Odoo Apps Store, or head to creyox.com to learn more. If your setup involves specific customization needs, the Creyox Technologies team is available to walk through configuration options and help tailor the module to your environment including which models to expose, how to structure masking rules, and which AI client to connect first.
Frequently Asked Questions
Yes. Every request passes through tool allowlists, IP restrictions, and field masking before it touches your data, and all actions still run through Odoo's native permission system.
Any MCP-compatible client can connect Claude, ChatGPT, Gemini, or others since Model Context Protocol is an open, vendor-neutral standard.
It's a standardized way for AI assistants to call external tools and access data, replacing fragile custom scripts with a structured, secure connection.
No. It adds an extra layer of control on top of Odoo's standard access rights and record rules nothing is overridden.
No. Installing the module and setting up a server profile doesn't require coding; day-to-day tasks like reports or dashboards are fully conversational.
Not unless you allow it. Sensitive fields can be masked automatically, so the AI never sees raw PII unless explicitly configured to.
Yes. Tool allowlists let you control exactly which actions read, write, or specific models an AI client is permitted to perform.
Yes. All interactions are recorded through the built-in audit log, so you always have a record of what was accessed or changed, and by whom.