odoo-mcp-server
A security-first MCP server that enables AI assistants to read and optionally write to Odoo 17 over XML-RPC, with model, field, and method allowlists to control access.
README
odoo-mcp-server
A minimal, security-first MCP server that lets AI assistants (Claude Code, Claude Desktop, or any MCP client) read and — optionally, behind explicit allowlists — write to an Odoo 17 instance over XML-RPC.
Built and used in production-adjacent work: MRP cost analysis, simulated purchase→manufacture→sale flows, and inventory cleanup driven entirely through an AI assistant, with the blast radius controlled server-side.
Design goals
- Tiny, auditable surface: one file (
server.py, ~450 lines), stdlib + the officialmcpSDK only. No third-party business logic, no telemetry. The only network destination is yourODOO_URL. - Read-only by default: write tools are not registered unless
ODOO_ENABLE_WRITES=1is set. There is no delete tool, ever. - Allowlists all the way down:
- models readable →
ODOO_ALLOWED_MODELS(supportsprefix.*wildcards) - models writable →
ODOO_WRITE_MODELS(subset of the above) - fields writable per model →
ODOO_WRITE_FIELDS - workflow methods callable →
ODOO_ALLOWED_METHODS(exactmodel:methodpairs)
- models readable →
- Defense in depth: these checks sit on top of the Odoo user's own access rights, which remain the primary boundary. See SECURITY.md for the full model.
Tools
| Tool | Registered | Purpose |
|---|---|---|
odoo_search_read |
always | Search + read records (domain, fields, limit, order) |
odoo_read |
always | Read records by id |
odoo_search_count |
always | Count records matching a domain |
odoo_fields_get |
always | Introspect a model's fields |
odoo_create |
writes enabled | Create a record (model + field allowlists enforced) |
odoo_write |
writes enabled | Update records (model + field allowlists enforced) |
odoo_call_method |
writes enabled + method list non-empty | Run an allowlisted workflow method (confirm a PO, validate a picking, apply an inventory count, …) |
Responses are hygienic by default: capped record counts, long/base64 values truncated, Odoo faults sanitized (no server tracebacks).
Installation
Requires Python 3.12+ and an Odoo 16/17 instance reachable over XML-RPC.
git clone https://github.com/rpl9717/odoo-mcp-server.git
cd odoo-mcp-server
python3 -m venv .venv
.venv/bin/pip install -r requirements.txt
cp .env.example .env
chmod 600 .env
# edit .env: ODOO_URL, ODOO_DB, ODOO_USER, ODOO_API_KEY, allowlists
In Odoo, create a dedicated user with a restricted access-rights group and an API key (Settings → Users → My profile → Account security). Don't run this with an admin key outside of a supervised test.
Sanity-check the configuration and connection:
.venv/bin/python server.py --check # auth + permission matrix
.venv/bin/python verify_mcp.py # end-to-end suite over real stdio
Hooking it up to Claude Code
.mcp.json in your project (no secrets here — credentials live in .env
next to server.py):
{
"mcpServers": {
"odoo": {
"command": "/path/to/odoo-mcp-server/.venv/bin/python",
"args": ["/path/to/odoo-mcp-server/server.py"]
}
}
}
One checkout can serve several databases: point ODOO_ENV_FILE at an
alternate env file per server entry.
Configuration reference
See .env.example — every variable is documented inline, including the reasoning behind the field-level allowlist entries.
Security
See SECURITY.md: the layered security model, why a custom
server instead of the community options, engineering notes (XML-RPC thread
safety, None-marshaling quirk, driving TransientModel wizards over RPC),
and the verification checklist.
License
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。