OdooForge MCP Server
OdooForge
AI-First ERP Configuration Engine for Odoo 18
CLI toolkit + MCP server that gives AI assistants complete control over Odoo 18. 79 tools. 6 skills. 4 specialist agents. Domain knowledge. Planning. Code generation.
Getting Started · Tool Reference · Architecture · Contributing
What Can It Do?
"I run a bakery with 3 locations and delivery — set up my Odoo"
"Show me how sales are doing this month compared to last month"
"Import my product catalog from this CSV file"
"Add a custom loyalty tier field to contacts as a selection"
"Generate a custom addon for recipe management with ingredient tracking"
"Something broke after that module install — help me fix it"
"Create an automation that sends a welcome email for new contacts"
"Run the restaurant recipe to set up a full POS system"OdooForge turns natural language into Odoo operations. It's not a one-time setup tool — it's an ongoing AI layer that helps you discover, build, populate, analyze, fix, and evolve your Odoo instance.
Quick Start
1. Install
pip install odooforge2. Initialize Workspace
mkdir my-odoo-project && cd my-odoo-project
odooforge initThis scaffolds everything you need:
.
├── CLAUDE.md # AI assistant context
├── .env # Connection settings (fill in your details)
├── .claude/
│ ├── skills/ # 6 Claude Code skills
│ │ ├── odoo-brainstorm/ # /odoo-brainstorm — explore ideas
│ │ ├── odoo-architect/ # /odoo-architect — design models
│ │ ├── odoo-debug/ # /odoo-debug — fix issues
│ │ ├── odoo-setup/ # /odoo-setup — deploy business
│ │ ├── odoo-data/ # /odoo-data — import & manage data
│ │ └── odoo-report/ # /odoo-report — dashboards & analytics
│ └── agents/ # 4 specialist subagents
│ ├── odoo-explorer.md # Read-only instance scout
│ ├── odoo-executor.md # Plan execution engine
│ ├── odoo-reviewer.md # Post-execution validator
│ └── odoo-analyst.md # Business data analyst
├── docker/
│ ├── docker-compose.yml # Odoo 18 + PostgreSQL 17
│ └── odoo.conf
├── addons/ # Your custom Odoo modules
├── .cursor/mcp.json # Cursor MCP config
├── .windsurf/mcp.json # Windsurf MCP config
└── .gitignore3. Start Odoo
# Edit .env with your connection details, then:
cd docker && docker compose up -d4. Start Coding
Open the workspace in your AI editor. The MCP configs are already set up — your assistant has access to all 79 OdooForge tools.
Ask your AI assistant to run
odoo_diagnostics_health_checkto verify everything is connected.
Updating Your Workspace
After upgrading OdooForge (pip install --upgrade odooforge), update workspace template files to the latest versions:
odooforge init --updateThis overwrites skills, agents, configs, and Docker files with the latest versions. Your .env is never overwritten.
The OdooForge Lifecycle
OdooForge isn't just for initial setup — it stays with you as your business evolves:
Discover → Plan → Build → Populate → Analyze → Fix → Evolve
↑ │
└────────────────────────────────────────────────────┘Phase | What Happens | Skills & Agents |
Discover | "What can Odoo do for my bakery?" |
|
Plan | "Here's a phased setup plan with 3 stages" | Planning tools |
Build | Install modules, custom fields, views, automations |
|
Populate | Import products, contacts, opening balances |
|
Analyze | "Show me revenue by location this month" |
|
Fix | "Why can't I create invoices?" |
|
Evolve | "We added delivery — what do we need?" | Back to Discover |
79 Tools Across 20 Categories
Category | # | Tools | Docs |
Instance | 5 |
| |
Database | 6 |
| |
Records | 6 |
| |
Snapshots | 4 |
| |
Modules | 6 |
| |
Models | 3 |
| |
Schema | 5 |
| |
Views | 5 |
| |
Reports | 6 |
| |
Automation | 5 |
| |
Network | 3 |
| |
Import | 3 |
| |
4 |
| ||
Settings | 4 |
| |
Knowledge | 3 |
| |
Recipes | 2 |
| |
Diagnostics | 1 |
| |
Planning | 3 |
| |
Workflows | 4 |
| |
Code Generation | 1 |
|
Domain Knowledge & AI Guidance
OdooForge includes built-in domain knowledge that helps AI assistants make informed decisions:
MCP Resources (6)
Structured knowledge accessible via odoo:// URIs:
Resource | Description |
| 35 Odoo 18 modules mapped to business language |
| 9 industry blueprints (restaurant, ecommerce, etc.) |
| Business terms → Odoo models/fields mapping |
| Naming conventions, field design, security patterns |
| Common customization patterns (trackable models, workflows) |
| Detailed blueprint for a specific industry |
MCP Prompts (4)
Guided workflows that provide step-by-step instructions:
Prompt | Description |
| Full business deployment from requirements |
| Custom feature creation with validation |
| Complete addon scaffolding |
| Systematic issue diagnosis and resolution |
Claude Code Skills (6)
Slash-command skills for Claude Code users (installed to .claude/skills/ by odooforge init):
Skill | Phase | Description |
| Discover | Explore customization ideas, discover modules, match blueprints |
| Plan | Design data models with naming conventions and security |
| Build | Full business deployment from natural language |
| Populate | Import, create, and manage business data |
| Analyze | Build dashboards, define KPIs, analyze business data |
| Fix | Diagnose issues with error mapping and snapshot rollback |
Specialist Agents (4)
Subagents that Claude dispatches for focused tasks (installed to .claude/agents/ by odooforge init):
Agent | Role | Description |
| Scout | Read-only instance discovery — modules, schema, customizations, data |
| Builder | Executes plans step-by-step with automatic snapshots before each mutation |
| Validator | Post-execution health checks, regression detection, verification |
| Analyst | SQL analytics, business insights, dashboard creation |
Agents have constrained tool access (principle of least privilege). Read-only agents can't write; the executor always snapshots before changes.
Industry Recipes
One-command setup for common business types:
Recipe | Modules | What It Sets Up |
Restaurant | POS, Kitchen, Inventory, HR | Table management, kitchen printing, food categories |
eCommerce | Website, Payments, Delivery, CRM | Online shop, cart, checkout, wishlists |
Manufacturing | MRP, Quality, Maintenance | Work centers, BoM, production planning |
Services | Project, Timesheets, CRM, Sales | Billable projects, task stages, invoicing |
Retail | POS, Inventory, Loyalty | Barcode scanning, stock alerts, loyalty programs |
"Run the restaurant recipe in dry-run mode first, then execute it"Configuration
OdooForge can be configured via environment variables. When running as an MCP server, pass these in your mcp.json or claude_desktop_config.json.
Variable | Description | Default |
| URL of the Odoo instance |
|
| Database to connect to by default |
|
| Admin username (for RPC) |
|
| Admin password |
|
| Master password (for DB creation) |
|
| PostgreSQL host |
|
| PostgreSQL port |
|
| PostgreSQL user |
|
| PostgreSQL password |
|
| Path to | Auto-detected |
MCP Client Config (mcp.json)
Option 1: Using uvx (Recommended)
This runs the latest published version of OdooForge isolated from your system.
{
"mcpServers": {
"odooforge": {
"command": "uvx",
"args": ["odooforge"],
"env": {
"ODOO_URL": "http://localhost:8069",
"ODOO_DEFAULT_DB": "odoo",
"ODOO_ADMIN_USER": "admin",
"ODOO_ADMIN_PASSWORD": "my_admin_password",
"ODOO_MASTER_PASSWORD": "my_master_password",
"POSTGRES_HOST": "localhost",
"POSTGRES_PORT": "5432",
"POSTGRES_USER": "odoo",
"POSTGRES_PASSWORD": "my_postgres_password"
}
}
}
}Option 2: Using Local Project (Development)
If you have customized OdooForge in a local virtual environment:
{
"mcpServers": {
"odooforge": {
"command": "/path/to/your/venv/bin/odooforge",
"args": [],
"env": {
"ODOO_ADMIN_PASSWORD": "secure_password"
}
}
}
}Architecture
graph TB
AI[AI Assistant<br/>Claude / Cursor / etc.] -->|MCP Protocol| MCP[OdooForge MCP Server<br/>79 tools · 6 resources · 4 prompts]
AI -->|Skills & Agents| CC[Claude Code Integration<br/>6 skills · 4 agents]
MCP --> Planning[Planning Layer]
MCP --> Workflows[Workflow Layer]
MCP --> Tools[Core Tool Layer]
MCP --> Knowledge[Knowledge Layer]
subgraph CC[Claude Code]
direction LR
Skills[6 Skills<br/>brainstorm · architect · setup<br/>data · report · debug]
Agents[4 Agents<br/>explorer · executor<br/>reviewer · analyst]
end
subgraph Planning[Layer 3: Planning]
direction LR
P1[Requirement Analysis]
P2[Solution Design]
P3[Plan Validation]
end
subgraph Workflows[Layer 2: Workflows]
direction LR
W1[Business Setup]
W2[Feature Builder]
W3[Dashboard]
W4[Integration]
end
subgraph Tools[Layer 0: Core Tools]
direction LR
T1[Records]
T2[Modules]
T3[Schema]
T4[Views]
T5[+ 13 more]
end
subgraph Knowledge[Layer 1: Domain Knowledge]
direction LR
K1[Module KB]
K2[Blueprints]
K3[Patterns]
end
subgraph Codegen[Code Generation]
direction LR
C1[Model Gen]
C2[View Gen]
C3[Security Gen]
end
Workflows --> Codegen
Tools --> Conn[Connections]
subgraph Conn[Connections]
direction LR
RPC[XML-RPC]
Docker[Docker]
PG[PostgreSQL]
end
Conn --> Odoo[Odoo 18<br/>Docker Container]
Conn --> DB[(PostgreSQL 17<br/>Docker Container)]
style AI fill:#5A67D8,color:#fff
style MCP fill:#714B67,color:#fff
style Odoo fill:#714B67,color:#fff
style DB fill:#336791,color:#fffsrc/odooforge/
├── cli.py # CLI — server, init, init --update
├── init.py # Workspace initializer — skills, agents, configs, Docker
├── server.py # MCP server — 79 tools, 6 resources, 4 prompts
├── config.py # Environment configuration
├── data/ # Bundled templates for odooforge init
│ ├── skills/ # 6 skill definitions (SKILL.md)
│ ├── agents/ # 4 agent definitions (.md)
│ └── ... # .env.example, docker, mcp configs
├── connections/
│ ├── docker_client.py # Docker Compose management
│ ├── xmlrpc_client.py # Odoo XML-RPC interface
│ └── pg_client.py # PostgreSQL direct connection
├── tools/ # One file per tool category (20 files)
├── knowledge/ # Domain knowledge (Layer 1)
├── planning/ # Planning engine (Layer 3)
├── workflows/ # Workflow orchestration (Layer 2)
├── codegen/ # Code generation engine
├── utils/ # Shared utilities
└── verification/ # Post-operation verificationSafety Features
OdooForge is designed to be safe for AI-driven operations:
Snapshots — Create backups before risky operations. Restore instantly. The
odoo-executoragent does this automatically.Confirmation guards — Destructive actions (delete, drop, uninstall) require
confirm=true.Namespace enforcement — Custom fields must start with
x_, custom models withx_. No accidental core modifications.Post-operation verification — Module installs, field creation, and view modifications are verified after execution.
Dry-run modes — Recipes and imports can be previewed before execution.
Input validation — Model names, field names, SQL queries, and domains are validated before execution.
Agent isolation — Read-only agents (explorer, reviewer, analyst) cannot modify your instance. The executor agent has constrained tool access.
Development
# Clone and install
git clone https://github.com/hamzatrq/odoo-forge.git
cd odooforge
uv sync --group dev
# Run tests (567 tests)
uv run pytest tests/ -v
# Run the server locally
uv run odooforgeSee CONTRIBUTING.md for detailed development guidelines.
Documentation
Document | Description |
Installation, workspace setup, first steps | |
Environment variables, Docker setup | |
All 79 tools with parameters and examples | |
Requirement analysis and solution design | |
Business setup, features, dashboards | |
Custom addon scaffolding | |
System design and data flow | |
Pre-built setups for common business types | |
Development setup and guidelines | |
Version history |
License
Latest Blog Posts
- Who's Calling? MCP Hosts Are an Identity Blind Spot (And the Spec Knows It)By Om-Shree-0709 on .mcpAgent IdentityOAuth 2.1
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/hamzatrq/odoo-forge'
If you have feedback or need assistance with the MCP directory API, please join our Discord server