In short: in Odoo MCP Pro you link your partner workspace to the workspace of each client. Your seat stays on your own bill, and one connection gives your AI the Odoo of every linked client. The client sees who from your team is in, can take people off, and can end the link without asking you.
Why is AI access to a client's Odoo awkward today?
Because the usual options work for the partner and leave the client in the dark. If you want your AI on a client's Odoo, you can:
- Use the client's credentials. You get the admin login or an API key and put it in your own setup. It works, but the client can't see who is using it, and ending it means replacing a key.
- Copy the client's Odoo into your own account. Now the instance exists twice, once with the client and once with you, and the two drift apart. We worked like this ourselves until 22 October.
- Join the client's workspace as a member. Then the client pays a seat for you, and a consultant with five clients has five seats on five bills.
Switching between clients was never the hard part. Local tools such as odoo-mcp-multi keep a profile per client in a config file on your laptop, and Odoo's own MCP server sits at /mcp on each database, one connection per client. Neither tells the client who is working in their Odoo. (We compared the options in every Odoo MCP server compared.)
What is a linked workspace?
A link joins two workspaces: yours as the implementation partner, and the client's as a company. The client's Odoo instances stay in the client's workspace, and you don't keep a copy.
Your administrators work in every linked company workspace. Your other members get in only where you add them. Inside the company workspace your people do the implementation work: they add Odoo instances, install apps and invite the client's own people. Billing stays with the client, and so do the client's vaults.
A company has at most one implementation partner. Two companies can't link to each other, and two partners can't either.
Can my whole team use it, each with their own login?
Yes. Everyone signs in to MCP Pro with their own account, and each person has their own connection to a client's Odoo, so there is no shared API key sitting in a config file. Your administrators reach every linked client. A consultant who only works for two of them gets added to those two. When someone leaves your company, you remove them from your workspace and they lose access to every client at once.
Who pays for the partner's seat?
The partner pays, once, in their own workspace. The company never pays for it.
| Pays the seat | Daily call limit used | Calls recorded at | |
|---|---|---|---|
| Partner consultant working in client A | Partner | Partner's plan | Client A |
| Same consultant working in client B | Partner (same seat) | Partner's plan | Client B |
| Client's own employee | Client | Client's plan | Client |
Each call is recorded against the client's workspace, so the audit trail stays with the client. The seat and the limit are yours.
What does the client see, and can they stop it?
Under Settings > Linked implementation partner the client sees everyone from your team who is in, with their role. The client's own administrators can take people off there, and either side can press Unlink. Your members then lose access to the workspace and its Odoo instances. The instances themselves stay where they are.
Nobody gets in without the client knowing. A request to an existing workspace goes to one of the client's administrators by email, and they answer it with Accept or Decline. Odoo's own rules still apply on every call, so the AI can do what the API key may do and nothing more. AI changes the speed, not the rules.
How does my AI switch between clients?
You tell it which client you mean. One MCP connection sees the Odoo instances of all your linked clients, each with the workspace it belongs to. A normal morning looks like this:
You: Which client Odoo instances can you reach?
AI: Six, in six workspaces. [list with workspace and label]
You: Show the overdue vendor bills at Client A.
You: Now the same at Client B.
You don't install a module on each client's Odoo, and you don't add a second connector per client. Your AI can manage the links as well: list your linked workspaces, send a link request, add or remove one of your members, and unlink after showing you what it will do. The one step it can't take is accepting a request. That is the client's decision, and they make it in the app.
How do I start?
Everything happens in the app:
- An administrator turns on Implementation partner under Workspace settings. It is in beta and off by default. Once it is on, the Linked workspaces page shows up in the sidebar.
- Choose Link workspace.
- If the client already uses MCP Pro, choose Link existing workspace and enter the email of one of their administrators. They get an email and accept under Settings.
- If the client is new, choose Create workspace. You are its administrator until the client's first administrator accepts your invite. From then on you are in it as their partner, and the client pays for their own people.
- Connect the client's Odoo instance in that workspace, and your AI can reach it straight away.
What does this not do yet?
Some things we kept simple on purpose. Access is per workspace, not per Odoo instance, so a client who wants separate environments for you needs two workspaces. A company has one partner; if a client works with two, one of them links. And Odoo logs the work under the user whose API key the connection uses. Our panel shows which person from your team made the call, while Odoo shows the API user.
We run our own client work this way
Pantalytics is an implementation partner too. Since 22 October six of our client workspaces have been linked to ours, and the duplicate copies of their Odoo instances in our own account are gone. Why partners use an MCP server for implementation work in the first place is in the best Odoo MCP connector for implementation partners.