Odoo Erp

A high-performance, secure Model Context Protocol (MCP) server built in Rust for seamless interaction with Odoo ERP systems. This server allows AI models (like Claude, GPT-4, etc.) to perform advanced ORM operations and data analysis on Odoo models natively.

Release License Platform

🚀 Features

• Standard CRUD: Search, Create, Update, and Delete any Odoo model.
• Advanced Tools:
  • search_count: Get record counts matching specific criteria.
  • read_group: Perform aggregated analysis (group by, sums, counts).
  • copy: Quickly duplicate existing records.
  • get_metadata: Inspect model structures and field definitions.
• Web Configuration UI: Modern, dark/light mode interface at http://localhost:3333 to manage multiple Odoo instances and custom AI prompts.
• Multi‑Instance Active Management: Each instance can be toggled active/inactive, and the server routes calls to the selected active instance.
• Global & Instance‑Specific Tool Permissions: Set a global default mode (CRUD or Read‑Only) and override per instance with its own mode or explicit allowed tools list.
• Protocol Version 2025-11-25: Built on the latest stable MCP specification for reliability and rich metadata support.

🛠 Tech Stack

• Rust: Language of choice for performance and safety.
• Tokio: Asynchronous runtime for non-blocking I/O.
• Axum: Lightweight web framework for the configuration UI.
• Reqwest: For handling Odoo JSON-RPC communication.
• Serde: For high-speed JSON serialization.

📦 Getting Started

Prerequisites

• Rust & Cargo (Stable)
• An Odoo Instance (v12 to v19 supported)

Local Development

1. Clone the repository:

   git clone <repo-url>
   cd odoo-erp-mcp
   

2. Run the server:

   cargo run
   

3. Configure via UI: Open http://localhost:3333 in your browser to add your Odoo instance credentials. These are saved securely in config.json.

MCP Communication Flow

sequenceDiagram
    participant AI as AI Model (e.g. Claude)
    participant Client as MCP Client (Antigravity/Cursor)
    participant Server as Odoo MCP Server (Rust)
    participant Odoo as Odoo ERP Instance

    AI->>Client: Send "Show me the top 5 customers"
    Client->>Server: tools/call (odoo-search-read)
    Server->>Odoo: JSON-RPC (search_read)
    Odoo-->>Server: JSON Result
    Server-->>Client: Tool Result (JSON)
    Client-->>AI: Result Context
    AI->>AI: Generate Human Response

🔌 Editor Integration

You can integrate this MCP server with various AI-powered code editors across different platforms (Windows, macOS, Linux). Ensure you have compiled the server or downloaded the executable.

Note on Paths: On Windows, ensure your path points to the .exe file using double backslashes (e.g., C:\\path\\to\\odoo-erp-mcp.exe) or forward slashes. On Linux/macOS, point to the generated binary directly (e.g., /path/to/odoo-erp-mcp).

Cursor

1. Go to Cursor Settings > Features > MCP Servers.
2. Click + Add new MCP server.
3. Set the Name to odoo-mcp.
4. Set the Type to command.
5. Set the Command to the absolute path of your odoo-erp-mcp executable.

Antigravity

Add the server to your mcp_config.json (typically located in ~/.gemini/antigravity/mcp_config.json or %USERPROFILE%\.gemini\antigravity\mcp_config.json):

{
  "mcpServers": {
    "odoo-mcp": {
      "command": "absolute/path/to/odoo-erp-mcp",
      "args": [],
      "env": {}
    }
  }
}

Visual Studio Code (via Extensions)

If you use MCP clients in VSCode (e.g., Cline or Claude Dev), edit your MCP settings file (e.g., cline_mcp_settings.json):

{
  "mcpServers": {
    "odoo-mcp": {
      "command": "absolute/path/to/odoo-erp-mcp",
      "args": [],
      "env": {}
    }
  }
}

PyCharm / JetBrains IDEs

For JetBrains IDEs using an MCP or AI assistant plugin that supports standard MCP servers:

1. Open the plugin's MCP server configuration.
2. Add a new Stdio based server.
3. Configure the command path to the odoo-erp-mcp executable.

🚢 Deployment & Release

Build for Release

To generate a highly optimized binary:

cargo build --release

The binary will be located at ./target/release/odoo-erp-mcp.

Deploying to Production

1. Copy the compiled binary and the src/index.html (if not embedded) to your server.
2. Ensure config.json is present or initialized via the UI on first run.
3. Add the binary path to your MCP client configuration (e.g. claude_desktop_config.json).

📄 License

This project is licensed under the MIT License.