AI Chats and LLM Odoo Integration. MCP v.19

Natural Language Business Intelligence (BI)

Instead of building complex pivot tables or custom SQL reports, query the database using your native language.

• Anomaly Detection: "Find all Sales Orders from last month where the margin is below 15%. Is this due to specific salesperson discounts or an increase in vendor costs"
• Trend Analysis: "Compare the performance of the 'Electronics' category in the USA between Q1 2025 and Q1 2026, adjusted for seasonal peaks"
• Ad-hoc Reporting: "Who are my top 5 customers in Germany who haven't placed an order in the last 60 days? Draft a follow-up email for each"

Intelligent Inventory and Supply Chain

AI can look past simple "min/max" rules to consider lead times, historical volatility, and external factors.

• Predictive Reordering: "Based on the last 6 months of Stock Moves and a 14-day vendor lead time, which products are at risk of stockout before the end of next month"
• Dead Stock Identification: "Identify items in WH/Stock that haven't moved in 120 days. Suggest a bundle deal involving these items and our current best-sellers"
• Warehouse Optimization: "Analyze our picking routes. Based on frequent together-buys, which products should be moved closer to the packing station"

Financial Automation and Smart Accounting

This goes beyond simple OCR (Optical Character Recognition); it involves semantic understanding of financial documents.

• Advanced Bill Matching: "Analyze this PDF invoice. Does it match the quantities and prices in Purchase Order #1234? If there is a discrepancy, highlight the specific line items"
• Cash Flow Forecasting: "Based on the Due Dates of all open Invoices and Bills, and our average customer payment delay, project our Citi bank balances for the next 30 days"
• Tax and Compliance Audit: "Scan all partner records for European entities. Flag any missing or mathematically invalid VAT numbers"

CRM and Hyper-Personalization

AI can synthesize data from Sales, Helpdesk, and Project modules to provide a 360-degree view.

• Customer Health Scoring: "Summarize the 'health' of Client X. Include recent sales volume, the number of open Helpdesk tickets, and their average time to pay invoices"
• Personalized Marketing: "Generate a promotional offer for customers who bought tables in Portugal in the last year, suggesting the new table protection cover. Write this in European Portuguese"
• Lead Triaging: "Score these 50 new leads based on their similarity to our most profitable accounts from 2026"

Automation

Apply routines and schedulers introduced, for example, by Claude:

• The "Daily Brief": 30 minutes before you start work. Claude reads all new Odoo Helpdesk tickets, labels them by priority, and writes a "TL;DR" summary in your Slack/Discord so you know exactly where to start.
• Predictive Stock "Early Warning": scheduled to run every Monday morning. Claude analyzes Stock Moves and current Sales Orders. If it predicts a stockout in the next 14 days based on vendor lead times, it sends a notification or creates a draft RFQ
• Automated Multi-Currency Reconciliation: runs daily during the night. The AI looks at "Unreconciled" entries. It uses reasoning to match complex bank statement text to Odoo Invoices, even if the reference numbers don't match perfectly, leaving only the "hard" ones for you to check.

Extract actionable meaning from the data you have! The list of AI applications is unlimited; it's totally up to you!

To ensure your Odoo data remains protected while interacting with LLMs, the module implements a multi-layer security architecture. Access is strictly controlled through dedicated roles, security deny rules (permission scopes), and specific safeguards for high-impact operations.

Mandatory User Roles

Connection to the MCP server is impossible without one of the two following security roles, ensuring that only authorized personnel can bridge Odoo with AI agents:

• MCP Server: User: The foundational role required for any AI interaction. Users must be assigned this group to allow AI agents to act on their behalf, enabling the retrieval of resources and the execution of authorized tools
• MCP Server: Administrator: Provides full authority over the system, including the management of OAuth clients, JWT signing keys, tool registries, and the high-fidelity audit log

Security Deny Rules and Scopes

The module introduces special security rules that allow defining what AI agents cannot do for specific users or user groups.

• Administrators can restrict specific users or groups in specific scopes, e.g., forbid reading a chatter or posting messages by AI
• If an AI agent attempts to perform the restricted action, the system would trigger an error, blocking the request and maintaining data integrity

Safeguards for Risky Operations

The Tool Registry allows for the classification of operations based on their potential impact on your business data

• Scopes that involve updating or deleting records are marked as "risky" within the system
• When a tool is flagged as risky, the AI agent is informed of the potential consequences. This allows the AI to proactively warn the user and wait for an explicit confirmation before executing the command, preventing accidental data loss or unintended bulk modifications
• Every high-risk execution is captured in the Audit Log, detailing the request and response payloads for full transparency.

Enterprise OAuth 2.0 Security

The module transforms Odoo into a dedicated OAuth Authorization Server. It manages the entire lifecycle of AI client security, including:

• Client Management: Register and configure "Public" or "Confidential" AI clients with specific redirect URIs
• Token Lifecycle: Secure handling of Authorization Codes and Refresh Tokens
• JWT Signing: Built-in infrastructure for JSON Web Token (JWT) signing using dedicated RSA keys
• Consent Management: User-facing consent screens to authorize specific AI access to Odoo data

High-Performance Infrastructure

Built for scale, the module includes specialized services to handle complex AI workflows:

• Transport Service: Manages multi-version protocol negotiation (Primary, Compatibility, and Legacy versions)
• Binary Service: Optimized handling of base64 data, mimetypes, and large file attachments for AI analysis
• Notification System: Real-time event broadcasting to AI clients via Odoo’s bus system
• Maintenance and Cleanup: Automated cron jobs to prune expired tokens, old audit logs, and temporary session data

MCP Catalog and Registry

The core of the server is a dynamic registry that allows administrators to define how AI interacts with Odoo:

• Tools: Define executable functions that AI can call to perform actions in Odoo, complete with JSON schema validation
• Resources: Provide AI with read-only access to Odoo data, documents, or system information
• Prompts: Create reusable prompt templates with arguments to guide AI behavior within your business context

Governance and Auditability

For enterprise compliance, the module provides a high-fidelity Audit Log system that captures every request and response payload. Administrators can:

• Monitor Sessions: Track active SSE (Server-Sent Events) and stream sessions in real-time
• Enforce Tool Rules: Apply granular "Tool Access Rules" to restrict which users or groups can execute specific AI functions
• Request Guard: Protect the system from malformed or unauthorized protocol requests.