HubSpot + Odoo integration

Odoo is a suite of business apps sharing one database, and a company that runs it for quoting, sales orders, invoicing and stock may well have its CRM app switched on too. The teams that reach us have chosen HubSpot for marketing and the sales pipeline and kept Odoo for everything after the order. That split holds until a seller needs to know what a customer has ordered, whether the last invoice was paid, or whether an item is in stock, and none of it is in the portal.

Two things make Odoo different from most ERPs we connect. It already has a CRM, with leads, opportunities and a pipeline, so the first decision is which system owns the pipeline rather than how to move data. Running opportunities in both is how a business ends up with two forecasts. And the edition and the plan decide what is possible. Odoo Community is open source, Odoo Enterprise is licensed, and Odoo Online runs the same software on Odoo’s own infrastructure. On Odoo’s hosted plans, the external API comes only with the Custom plan.

HubSpot publishes an Odoo app of its own, and for a team whose question stops at companies, contacts and the pipeline, it is the cheapest answer. The request that needs a build is the one that crosses into orders and invoices: show open sales orders on the company, flag the customer with an overdue invoice, create the Odoo customer when a deal closes.

Five objects carry most of the value, as the table lists. Odoo keeps companies and people in one contact model, and its external API tells them apart with a company flag, so the load splits one Odoo model into HubSpot companies and HubSpot contacts. Sales orders become deals, customer invoices land on a custom object associated to the order, and products become HubSpot products.

Every record keys on its Odoo id, held in a custom HubSpot property. Names are the wrong key here for a specific reason: HubSpot’s own data sync matches companies on name or domain, and it will pair two records that share a name even when their domains differ. Contacts are the exception, where email is the practical key and the Odoo id breaks ties between people who share an address.

Direction follows ownership. Odoo owns the order, the invoice and the product, so those move one way and stay read only in the portal. HubSpot owns the pipeline. The one write back worth building early is a closed won deal creating or updating the Odoo customer, and we hold even that behind a review step for the first month.

The shape is a polling watermark sync. Odoo stamps a write_date on every record of a model that keeps access log fields, which is the default, so a service can ask each model for the records changed since its last run. It reads with search_read, which searches and reads in one call and one transaction, and it keeps a watermark per model so a restart resumes rather than reloads.

On Odoo 19 and later the service calls the JSON-2 API: a POST to /json/2/ with the model and method in the path and an API key sent as a bearer token. The key belongs to a dedicated bot user, so its access rights limit what the integration can touch. Earlier versions use the XML-RPC object service, which takes the database, user id and password on every call.

Writes into HubSpot are batch upserts keyed on the Odoo id, then a separate association pass, so a replay changes nothing. Where speed matters, an Odoo automation rule can send a webhook notification when a record changes, and the service treats it as a prompt to read early rather than as the data itself. A reconciliation endpoint compares counts per model and reports the difference instead of correcting it quietly.

What decides the route is how far past the CRM records you need to go. HubSpot’s Odoo app syncs companies, contacts, leads, opportunities, products and activities, one way or both, with default field mappings and a filter on any Odoo field or HubSpot list. The app itself is free, though it needs an Odoo subscription, custom field mappings need Data Hub Starter or above, and it does not carry sales orders or invoices.

Once orders and invoices are in scope, the choice is a custom API integration or an iPaaS, and both need Odoo’s external API, which on the hosted plans means the Custom plan. Connectors sold as modules on the Odoo Apps Store are a further option only away from Odoo Online, which Odoo documents as incompatible with Apps Store modules, so they need Odoo.sh or your own server.

1. Scope. We settle which system owns the pipeline, which Odoo edition, version and plan the database runs, and whether HubSpot’s own app already covers the need.
2. Architect. We fix the models, the Odoo id keys, the watermark per model, the archived record rule and the association order, and write them down.
3. Build. The service is written against a duplicate database and a HubSpot sandbox, with a dedicated bot user and a portal allowlist from the first commit.
4. Validate. We run a full reload, reconcile counts per model against Odoo, and walk the differences with whoever owns the ledger.
5. Operate. Cutover is a key swap. After that we rotate keys before they expire, watch the run history, and retest after every Odoo upgrade notice.