Ask Anything.
Do Anything in Odoo.
Safely.
Niyu Labs MCP Server turns Odoo into a governed AI workspace. Connect modern AI clients to real Odoo data, expose only approved apps and models, choose the execution user, control CRUD access, inspect diagnostics, and audit every action from one Odoo-native control layer.
AI without governance is risky.
Odoo without context is confusing.
Business language to real Odoo models
Customers, vendors, RFQs, invoices, receipts, deliveries, opportunities, products, warehouses, and custom workflows live across many technical models. The module gives AI a discoverable Odoo guide, glossary, model guidance, prompts, and resources.
Access is capped from three sides
Final MCP permission is not random. It is limited by normal Odoo ACLs, the selected execution user, and the assigned access bundle. Admins control exactly what AI clients can read, create, update, or delete.
Made for real administrators
Admins do not need to manually guess every Odoo model. Choose business apps such as Inventory, Sales, Purchase, CRM, Accounting, Project, Manufacturing, Website, Helpdesk, HR, and Contacts. Fine-tune only when needed.
Turn MCP on.
Control everything from Odoo.
Enable the server, generate the bearer token, choose allowed browser origins, select the execution user, copy Cursor configs, configure rate limits, set IP allowlists, manage log retention, and add an instance playbook for your business rules.
Endpoint + token
Native `/mcp` endpoint with Bearer Token mode for production clients and generated connection snippets.
Execution user
All MCP tool calls run as the selected internal Odoo user, so permissions remain predictable.
Allowed origins + IPs
Restrict browser origins and client IPs for safer hosted MCP access.
Playbook for AI
Describe custom rules, modules, field meanings, naming conventions, and workflows so AI understands your instance.
Expose apps, not chaos.
Then fine-tune models.
Pick business apps. The module computes matching Odoo models automatically. Add manual include models for special workflows. Add exclude models when a model must never be visible to MCP clients.
Include and exclude models
Add one-off models when your workflow needs them. Exclude risky or irrelevant models even if they match an app pattern.
See model counts before using AI
Admins can see which bundles are active, how many models are matched, and whether access is broad enough for the selected AI workflow.
Read-only by default.
CRUD only when allowed.
This is not a blind gateway. Create, update, and delete tools appear only when the selected execution user, Odoo ACLs, and bundle permissions allow them.
Odoo ACLs
Normal Odoo user permissions still apply.
Execution user cap
The selected user has Read, Write, Create, Delete flags.
Bundle cap
Each access bundle has its own permission preset.
Final MCP tools
AI sees only the tools that survive all caps.
Give AI useful tools.
Not unlimited power.
The module exposes safe discovery tools, model tools, CRUD tools, prompts, and resources. Availability changes based on your actual permissions.
Understand before acting
Health, diagnostics, system info, instance guide, exposed model list, semantic glossary, business question resolver, search, and fetch.
odoo_instance_guide
odoo_list_models
odoo_resolve_question
odoo_semantic_glossary
Read and act on Odoo records
Search records, read IDs, browse with pagination, count records, inspect model fields, create records, update records, delete records, and create contacts when allowed.
odoo_browse / odoo_count
odoo_create / odoo_write
odoo_delete
create_res_partner_contact
Built-in guidance for AI clients
Clients can read Odoo resources and use prompt templates for safe reads, safe updates, inventory checks, and sales assistant workflows.
odoo://models
odoo://health
odoo://diagnostics
safe_odoo_read / safe_odoo_update
One Odoo server.
Many AI clients.
Use the same governed MCP layer from ChatGPT, Claude, Cursor, Codex, VS Code, and other compatible tools. Remote HTTP for hosted environments. Local STDIO bridge for development and localhost workflows.
Connect with ChatGPT
Paste the endpoint, configure authentication as supported by the client, and ask real Odoo questions using the selected execution user.
Connect with Claude
Use Claude with the same permissioned Odoo endpoint, resources, prompts, and model restrictions.
Copy config from Odoo
Use generated remote config for hosted Odoo, or local STDIO bridge config for development environments where the AI client runs on your machine.
AI understands Odoo words.
Not only table names.
The module gives clients business vocabulary and model hints so questions like âoverdue invoicesâ, âRFQsâ, âinternal transfersâ, âstock by warehouseâ, and âstale opportunitiesâ map to the right Odoo models.
Business glossary
Customer/vendor/contact to res.partner. Invoice/bill to account.move. RFQ to purchase.order. Transfer to stock.picking.
Instance playbook
Add your custom business rules, custom modules, field meanings, naming conventions, and approval flows.
Question resolver
AI can resolve business-language questions into likely models, fields, domains, ordering, and ambiguity notes.
Model guidance
Per-model business terms, prompt guidance, cache TTL, and safe field metadata help clients work with fewer mistakes.
Know what is exposed.
Know what happened.
The module is built for production support: health checks, exposed model counts, execution-user visibility, auth status, security settings, audit logs, and automatic cleanup.
Health and diagnostics
Check enabled state, auth mode, bearer token, execution user, exposed model count, origin restrictions, generic method-call status, and client configs.
Audit every tool call
Track user, model, operation, method, record IDs, record count, request data, status, error message, IP address, user agent, duration, and timestamp.
Sensitive-field filtering
Passwords, tokens, API keys, secrets, sessions, OTP fields, payment-card-like fields, and bank-account-like fields are blocked from exposure.
No blind all-field reads
`__all__` field reads are blocked. Clients must request explicit fields or use safe defaults.
Forced-domain enforcement
Forced domains are applied to reads and mutations, so scoped access remains scoped.
Rate limit + IP allowlist
Limit requests per minute and restrict clients by IP when needed.
What customers see inside Odoo
Authentication and endpoint setup
Clear endpoint, token, and client setup experience for AI tools.
Manual overrides
Include special models and exclude models that should stay hidden.
Full admin control
Enable MCP, choose execution user, control auth, origins, IPs, cache, logs, and configs.
Operational audit trail
Inspect successful, denied, and failed MCP activity.
Let AI work with the ERP.
Without opening the whole ERP.
Examples of what teams can ask when the right apps, models, and permissions are exposed.
Sales
âShow quotations older than 14 days.â
âWhich customers have open follow-ups?â
âCreate a draft note for this opportunity.â
Inventory
âCheck stock by warehouse.â
âFind pending receipts.â
âExplain stock moves for this product.â
Accounting
âList overdue invoices.â
âFind unpaid vendor bills.â
âSummarize invoices by customer.â
Purchase
âShow open RFQs.â
âFind purchases from this vendor.â
âCreate a draft vendor contact.â
Everything customers ask before buying
Does this work only with ChatGPT?
No. It is built for MCP-compatible clients such as ChatGPT, Claude, Cursor, Codex, VS Code, and other tools that can connect to MCP servers.
Can it create, update, or delete Odoo records?
Yes, when allowed. CRUD access depends on the execution user, normal Odoo ACLs, and the access bundle permission preset. New setups are read-only by default.
Do admins need to know technical model names?
No. Built-in access bundles cover common apps such as Inventory, Sales, Purchase, CRM, Accounting, Project, Manufacturing, Website, Helpdesk, HR, and Contacts. Technical model overrides are available only when needed.
What keeps AI from accessing everything?
Execution user permissions, assigned bundles, Odoo ACLs, record rules, permission presets, sensitive-field filtering, blocked models, allowed origins, IP allowlists, and audit logs.
Does it support Cursor?
Yes. The module provides Cursor-ready remote MCP configuration and a local STDIO bridge configuration for development or localhost Odoo instances.
Can it call arbitrary Odoo methods?
The generic method-call endpoint is disabled by default for safety. Production workflows should use explicit MCP tools and approved model permissions.
Your Odoo becomes
AI-ready, permission-safe, and auditable.
Native MCP endpoint. Bearer authentication. Execution-user control. App-based access bundles. Include/exclude models. Safe CRUD permissions. Prompts and resources. Semantic glossary. Diagnostics. Audit trail. One serious MCP layer for Odoo teams.
|
NIYU LABS
Secure AI infrastructure for Odoo teams
|
Queries & Support
info@niyulabs.com
|
Please log in to comment on this module