Odoo news
Odoo.sh AI means two things in 2026: since April 2, 2026, developers can run Claude Code and OpenCode inside Odoo.sh, and Odoo 19 Enterprise, which Odoo.sh hosts, ships a dedicated AI app with agents, Ask AI search, AI fields, AI server actions and voice transcription. Here is what each one does, what it costs and what you need.
What if your ERP could draft its own emails, answer your team's questions, score your leads, transcribe your meetings, and help your developers write code, all without adding a separate app? With Odoo 19 on Odoo.sh, most of that is available today.
Odoo has been building AI into the core of its ERP. With Odoo 19 released in September 2025 and AI coding tools added to Odoo.sh in April 2026, this is the moment to understand what is actually available, and what it means for your business and your team.
This is not a feature list. It is a practical guide to what AI on Odoo.sh looks like in real use, written for both the developers building on the platform and the business owners running their operations on it.
Odoo.sh is Odoo's official cloud platform: it manages your hosting, staging branches, automated builds, backups and deployments, so you can focus on what matters. Now it is also where Odoo's AI features and AI coding tools come together.
What is Odoo.sh AI, and why does it matter now?
Odoo.sh AI is the AI that reaches teams on Odoo.sh: the AI app of Odoo 19 Enterprise for business users, and Claude Code and OpenCode in the Odoo.sh development environment for developers. Odoo built the business side into its core apps rather than selling it as a separate product.
Odoo built AI directly into the core system and shipped it as part of Odoo 19 Enterprise: no separate license and no third-party app to maintain. On Odoo.sh and on-premise you connect your own OpenAI or Google Gemini key, and the provider bills you for what you use.
The result is a system where a sales manager, an accountant, an HR officer and a developer can all benefit from AI inside the same platform, on the same day, without anyone needing to become an AI expert first.
- 19.0the first Odoo version with the AI app on Odoo.sh
- 1IAP credit per document for bill digitization
- 2AI providers in Odoo 19: OpenAI and Google Gemini
- 2coding agents built into Odoo.sh: Claude Code and OpenCode
For business users: what AI actually does inside your Odoo
Here is what changed in Odoo 19 for the people running day-to-day operations, and where each feature saves time.
Your team gets an assistant that knows your business
Odoo 19 introduces configurable AI agents: you set each one's prompt, model, response style and topics, and point it at your own sources, such as PDFs, web pages, files in Documents and Knowledge articles. The agent looks those sources up when it answers, so your team or your customers get replies based on your catalog, policies or procedures.
Agents act through tools, which are server actions. Out of the box they can search and open views, retrieve information and create leads; other actions need tools you add. Assign an agent to a website Live Chat channel and it answers support questions around the clock, or keep one internal so your HR team stops answering the same leave-policy question every week.
Find any record by simply asking for it
How often does someone on your team spend minutes building filters to find something they could describe in one sentence? Ask AI, opened with Ctrl+K, the AI button or the search panel, turns a plain-language request such as "overdue invoices from last month" into an Odoo filter and opens the matching list.
Build automations by describing what you want
Odoo's automation engine is powerful, but configuring it has always needed some technical know-how. Odoo 19 adds AI server actions: you still set up the action, but its logic is a prompt, and when it runs the AI decides which tool to call or which value to write. "When a sale order is confirmed, draft a thank-you note for the customer" becomes a prompt instead of code.
Never lose what was decided in a meeting
Type /Voice Transcript in the notes or description of a CRM opportunity, a project task, a calendar event or a Knowledge article to record a meeting. Odoo writes a live transcript and an AI summary into that field and discards the audio, so your sales team spends less time writing call notes and your project manager knows what was agreed.
Vendor bills that read themselves
Document digitization reads vendor bills, receipts and scanned documents with OCR and AI, at one IAP credit per document. It looks for a matching purchase order and links the order lines when the reference and the amounts match, so your accounts team reviews and approves instead of typing.
CRM that tells you which leads are worth your time
Predictive lead scoring, in Odoo since version 13, estimates each lead's chance of winning from your own won and lost leads. Odoo 19 shows the main factors behind that figure, and Ask AI can suggest next steps and draft the follow-up email, so your team starts each week on the leads most likely to close.
AI in Odoo 19 is both a dedicated app and a set of features inside the apps your team already uses: CRM, Accounting, Documents, Knowledge, Live Chat and more.
Which AI engine powers Odoo AI, and can you choose?
Yes, between two providers. Odoo 19 works with OpenAI and Google Gemini; each agent has its own model, while some features always use OpenAI. Models change with Odoo updates, so the list below reflects the Odoo 19 code in October 2026.
| Provider | Models | Used for |
|---|---|---|
| OpenAI | GPT-4o, GPT-4.1, GPT-4.1 Mini, GPT-5, GPT-5 Mini and ChatGPT (Auto) | Agents, and always for AI fields, AI server actions and voice transcription |
| Google Gemini | Gemini (Auto) | Agents that you set to a Gemini model |
| Custom or self-hosted model | Not supported | Odoo 19 calls OpenAI and Google directly; there is no setting for another endpoint |
Sources: Odoo's AI API keys documentation and the AI modules of Odoo 19 Enterprise. Now for the developers: Odoo.sh itself has AI built into the development environment, and it changes the way Odoo modules get built.
Can you use Claude Code with Odoo.sh?
Yes. Since April 2, 2026, Odoo.sh includes Claude Code and OpenCode in its development environment, and an opt-in writable deploy key lets you commit from Odoo.sh terminals. Other tools, such as Cursor or GitHub Copilot, work on your local copy of the Odoo.sh repository.
If you have used AI for Odoo development before, you know the frustration. Open Claude or ChatGPT, paste a chunk of Python, spend two paragraphs explaining your _inherit chain and custom field names, receive a response that largely ignores that context, copy it back to your editor, hit an error, and start the loop again. The AI never actually sees your project: you are the bottleneck, shuttling context between two disconnected worlds.
Running the agent inside Odoo.sh closes that gap. Claude Code works in your build's terminal, next to your module structure, your __manifest__.py and your running container. Ask it to review your sale.order extension and suggest test cases, and it reads your actual overrides instead of guessing. The Odoo.sh changelog has the details.
Works with the tools developers already use
Claude Code and OpenCode run inside Odoo.sh; Cursor, Gemini Code Assist and GitHub Copilot work on a local clone of the same repository, and Odoo.sh builds the branches you push. Odoo Experience 2025 in Brussels featured a partner session, "Beyond Code Generation: Integrating AI into Odoo's Development Lifecycle - Lessons Learned", on what works, what breaks and where Odoo-specific challenges like version confusion, ORM nuances and XML view inheritance trip these tools up. The takeaway: AI amplifies Odoo expertise, and a developer who understands the framework gets much better results than one who does not.
AI fields mean fewer custom compute methods
How often do you write an @api.depends compute method just to summarize field values into a text field? The AI field type in Studio handles these cases with a prompt: define what you want generated, reference the source fields, done. It is faster to build, and your client can read and understand the configuration.
Prompt-based server actions reduce client delivery boilerplate
Client customizations repeat the same patterns constantly: on confirmation update this field, when the status changes send this notification. With AI server actions, the rule's logic is a prompt instead of Python, which means less boilerplate per project and fewer scripts to maintain six months later.
MCP: the pattern worth knowing now
The Model Context Protocol (MCP) is the bridge between AI assistants such as Claude or ChatGPT and Odoo's data. An MCP server receives a plain-language request from the assistant, runs the matching Odoo call and returns structured results. The TeqStars Advanced MCP Server does this inside Odoo 17.0 to 20.0, Community and Enterprise, with approvals and an audit log. If you build your own, use the JSON-2 API: Odoo has scheduled XML-RPC and JSON-RPC for removal in Odoo 22. Our guide to connecting Claude to Odoo walks through it, Claude Code included.
Same feature, two very different perspectives
Every AI capability in Odoo 19 on Odoo.sh means something different depending on whether you build with it or use it every day:
| Feature | What it means for developers | What it means for business users |
|---|---|---|
| AI agents | Configure once, add sources and tools, assign to Live Chat: no Python for the standard cases | Answers on products, invoices and HR policy, around the clock |
| AI fields in Studio | Replace @api.depends compute methods written only to generate text | Records fill descriptions, summaries and notes from a prompt |
| AI server actions | Fewer bespoke automation scripts to write and maintain per client | Automations whose logic is described in plain language |
| Ask AI search | Fewer custom filters needed in list views | Find any record by asking a plain-English question |
| Claude Code on Odoo.sh | The AI sees your actual addon code: no more context switching | Faster, more accurate customization from your Odoo partner |
| Voice transcription | Meeting summaries land on the task, cutting admin on project delivery | Decisions and action items captured on the CRM record or task |
| Bill digitization | No custom field-mapping scripts for standard vendor bills | Vendor bills read and matched to purchase orders |
What changed for AI in Odoo 20?
Odoo 20, unveiled at Odoo Experience in Brussels on September 24, 2026, extends the same AI app and changes how it is paid for: AI now runs on Odoo IAP credits instead of your own provider key. The main AI items in the Odoo 20 release notes:
Agents that act
Agents can now create and update records, ask about a file, send files and generate images.
Agentic automation
Automations can hand a multi-step task to an agent, with a confirmation when it reaches its tool call limit.
Automatic model selection
Choose a provider, and Odoo picks the model for each task.
Voice interaction
Talk to an agent, alongside stored conversations and interactive replies with buttons.
Model-bound agents and skills
Agents tied to a specific model, and topics renamed to skills.
AI in more apps
Financial insights in Accounting, AI renaming in Documents, similar-ticket detection in Helpdesk and a Website assistant.
For the rest of the release, read our Odoo 20 release overview.
What do you actually need to get started?
Five things for Odoo 19, and no surprises:
- Odoo Enterprise: the AI app, agents, AI fields and voice transcription are Enterprise-only, and Odoo.sh runs on the Custom plan.
- Odoo 19 or later: the AI app first shipped in Odoo Online 18.3, which never ran on Odoo.sh or on-premise.
- An Odoo.sh subscription, for managed hosting, staging branches, automated builds and the Claude Code and OpenCode tooling.
- Your own API keys on Odoo.sh: OpenAI for AI fields, AI server actions and voice transcription, and OpenAI or Gemini for agents.
- On your own server, the PostgreSQL pgvector extension, which the AI module needs to install.
5 things to do on your Odoo.sh instance this week
Install the AI app and add your keys
Install AI from Apps in Odoo 19, then add your OpenAI or Gemini key in its settings. Features across the apps become available as soon as a key is in place.
Set up one AI agent for your biggest support pain point
Give it your product catalog, FAQ or pricing policy as sources and assign it to Live Chat. Measure how many questions it answers before they reach your team.
Add an AI field to one high-volume form
Product descriptions and lead summaries are the easiest wins. Write the prompt, reference the relevant fields, and let records fill themselves.
Replace one repetitive manual step with an AI server action
Pick the task someone does by hand after every sale confirmation or invoice approval, describe its logic in a prompt, and test it on a staging branch.
Developers: try Claude Code in an Odoo.sh terminal
Open a development branch, start Claude Code next to your modules, and use it on your next feature. Compare delivery time with your last module.
Frequently asked questions
Is Odoo AI available on the Community edition?
No. The AI app, agents, AI fields, AI server actions and voice transcription are part of Odoo Enterprise. Community keeps predictive lead scoring and AI text generation in the editor, but has no AI app; on Odoo.sh you also need the Custom plan.
Do I need to purchase extra AI tokens or credits?
In Odoo 19 on Odoo.sh or on-premise, you add your own OpenAI or Gemini key and pay that provider; on Odoo Online, AI works without your own key. Bill digitization uses one IAP credit per document. From Odoo 20, AI features run on Odoo IAP credits.
Can you use Claude Code with Odoo.sh?
Yes. Since April 2, 2026, Claude Code and OpenCode run inside the Odoo.sh development environment, and an opt-in writable deploy key lets you commit from its terminals. Cursor and GitHub Copilot work on a local clone of your Odoo.sh repository instead.
What is vibe coding in Odoo.sh?
Vibe coding is AI-assisted development: you describe what you want in plain language and an AI coding agent writes the code. On Odoo.sh it means running Claude Code or OpenCode inside your project, with your modules and running container in reach, rather than in a disconnected browser tab.
Which Odoo version do I need for the full AI feature set?
Odoo 19 for the AI app on Odoo.sh or on-premise; Odoo 18.3 had it only on Odoo Online. Odoo 20 adds agents that create and update records, voice interaction and automatic model selection. Odoo 16 and 17 had bill digitization, lead scoring and, in 17, AI text in the editor, but no AI app.
Can I use my own AI model instead of ChatGPT or Gemini?
Not in Odoo 19. Its AI calls OpenAI or Google Gemini directly, and there is no setting for a custom endpoint or a local model. Teams that need another model add it through custom development or a third-party module.
Does Odoo AI work on on-premise installs or only on Odoo.sh?
Both, and on Odoo Online. Odoo 19 AI works on any Enterprise deployment; on Odoo.sh and on-premise you add your own provider key, and on-premise also needs the pgvector extension. Odoo.sh adds staging branches, automated builds and the AI coding tools.
Which AI features came with Odoo 19?
A dedicated AI app with configurable agents, Ask AI search, AI fields in Studio, AI server actions, voice transcription with summaries, and AI help in apps such as CRM and Documents.
The bottom line
Odoo 19 on Odoo.sh brought AI from a feature in one module to a layer across the business: answering questions, drafting communications, reading documents, explaining lead scores and helping developers write code. Odoo 20 builds on it with agents that act and AI paid through Odoo credits.
Business users get time back from tasks they should never have done by hand. Developers get an assistant that sees the codebase they are working in. Both work inside the same platform, without stitching separate tools together.
The gap between using Odoo and using Odoo intelligently is smaller than it has ever been.
The features are already in your instance. The advantage is simply in switching them on.
If you are already on Odoo.sh and Odoo 19, the question is not whether to use AI. It is which process you improve first.
How current this article is
AI features described here require Odoo Enterprise, with the AI app on Odoo.sh from Odoo 19. Models and pricing change with Odoo updates; check the Odoo 19 release notes, the Odoo.sh changelog and the Odoo documentation for the latest details.