Teams running Odoo constantly hit the same wall: the data they need lives in the ORM, but turning it into a purpose-built screen, a compute-heavy dashboard, or a small automation means a developer round-trip every time. Odoo native gives you list/form/kanban/pivot/graph views and the studio-style field builder, but the moment you want a bespoke React-driven interface, a chart type the built-in graph view does not render, or an AI assistant that can read your models and draft the models.Model + ir.model.access.csv + view XML for a new feature, you are back to hand-coding or waiting in the backlog. That gap — between "the data is in Odoo" and "the interface my team actually needs" — is what this build-to-order tool is designed to close.

We build MCP Studio as a governed bridge between AI agents and your Odoo instance. At its core is a Remote MCP (Model Context Protocol) server we implement against your database: it exposes a curated, permissioned set of tools — read model metadata, query records through the ORM (never raw SQL), draft a module skeleton, or push a generated OWL/React component — so an AI agent can operate on your data within boundaries you approve rather than with open write access. Around that we deliver a builder surface where a prompt like "show open sale orders by salesperson as a stacked bar with a drill-down to margin" is translated into a working EChart-based visualization embedded in a QWeb/OWL view, backed by a real Odoo model or a computed dataset. Generated artifacts are genuine Odoo code — a proper __manifest__.py, models with fields and @api.depends compute methods, security via ir.model.access.csv and record rules, XML/OWL views, and where relevant automated actions and server actions — so what the agent produces is reviewable, versionable, and installable, not a black-box widget.

Technically, the tool is engineered for safety and auditability because it sits close to your data. The MCP layer authenticates against a dedicated Odoo API user scoped by groups and record rules, so agent access inherits the same access-control model as any human user; all agent-initiated reads and writes go through the XML-RPC/JSON-RPC API or an internal service layer we control, are rate-limited, and are logged for review. The React/EChart builder outputs components that mount inside Odoo's web client (OWL) or as embeddable dashboards, with chart definitions stored as data so non-developers can tweak series, filters, and grouping without a redeploy. We support Odoo 17.0, 18.0 and 19.0, and we scope Community vs Enterprise explicitly up front — for example, whether your target relies on Enterprise-only apps, the enterprise web client theming, or studio artifacts — so the deliverable installs cleanly on the edition you actually run.

Because this is build-to-order, delivery follows a defined path: a scoping call to fix the model surface, the agent tool boundaries, and the dashboards/automations in scope; a written scope and fixed quote; development on a staging instance; UAT with your team; then production install with a rollback plan and a support window. Typical delivery is one working week from confirmed scope, depending on how many models the agent must reach, how many bespoke views/charts you need, and your security review requirements. Pricing starts from $999 (indicative, single-company base scope); wider agent tool coverage, additional external integrations wired through MCP, deeper localization, and larger data-migration or multi-company volume increase the quoted scope. The from-price is indicative only — you receive a firm fixed quote after the scoping call, never a surprise on the invoice.