MCP (Model Context Protocol) is an open standard that lets an AI model connect to external tools and data through one consistent interface. Anthropic introduced and open-sourced it in November 2024, and it has since been adopted well beyond Claude. When the tool on the other end is your Odoo, MCP is what lets Claude read and act on your real business data — without a bespoke integration for every question.
The problem MCP solves
Before MCP, every AI-to-software connection was custom. Want Claude to see your Odoo invoices? Someone writes a specific integration. Want it to also read your CRM, or your stock, or trigger an action? More custom code, each with its own auth, its own data shapes, its own maintenance. This is the “N×M problem”: N AI clients times M tools, each pair wired by hand.
MCP replaces that with a single protocol. A tool exposes its capabilities once, in a standard way; any MCP-capable AI client can then use them. Anthropic describes it as “a universal, open standard for connecting AI systems with data sources, replacing fragmented integrations with a single protocol.”
The three things MCP standardises
An MCP server (the tool side) can offer three kinds of capability, and it’s worth knowing the vocabulary:
- Tools — actions the AI can call, like “search records” or “create an invoice”. Each has a name, a description, and a typed input schema so the model knows how to call it.
- Resources — data the AI can read, addressed by URI (a document, a record, a file).
- Prompts — reusable, parameterised prompt templates the server can offer.
The AI client discovers these at connection time. That discovery step is the magic: you don’t tell Claude what your tools are — it asks, and adapts.
Why this matters specifically for Odoo
Odoo is a database of everything a business runs on — sales, inventory, accounting, projects, HR. It already has a mature permission system (ACLs, record rules, field-level rights) and an ORM. What it lacked was a standard way for an AI to reach that data safely.
An MCP server inside Odoo closes that gap. Instead of exporting data to some external AI platform, the AI connects to Odoo directly and every call runs through Odoo’s own access control. The result:
- No data leaves your instance to a third-party AI connector.
- The AI is bounded by the user’s Odoo rights — it can only see and do what that user can.
- One integration, all of Odoo — the same protocol reaches any model the user is allowed to touch, not a hand-picked subset.
That is exactly what Cledoo MCP does: it turns your Odoo into a native MCP
server, so Claude (or any MCP client) connects to /mcp and works under your
existing permissions.
MCP is a standard, not a Claude feature
A common misconception: MCP is “an Anthropic/Claude thing”. It started at Anthropic, but it was open-sourced as a specification with SDKs, and the wider ecosystem — other AI clients and thousands of community servers — has adopted it. Building on MCP means you’re building on an open protocol, not a single vendor’s API. Your Odoo MCP server works with any client that speaks MCP, today and as the standard evolves.
The takeaway
MCP is the “USB-C for AI tools”: one standard port instead of a different cable for every device. For Odoo, it means an AI can finally reach your business data through a single, permission-aware interface that lives inside your own instance — which is the difference between a demo and something you’d actually run in production.