Loomworks

Proof / Case study

Internal platform build. No client engagement.

Create records, run reports, and trigger workflows in Odoo from chat.

Loomworks built a working chat interface for creating Odoo records, running reports, and triggering workflows, with tenant access and credentials kept separate.

Built and operated by Loomworks as an internal platform. The evidence below covers shipped software and test results. No client outcome is claimed.

Industry
Internal platform (AI + Odoo tooling)
Service
Custom Module Development / AI Integration

01

Starting point

Odoo covers most day-to-day ERP operations well, but non-technical users still have to navigate module-specific screens to do routine work: creating records, running reports, triggering workflows.

The goal was to prove that a conversational layer could sit on top of Odoo and handle real operations safely, with proper multi-tenant isolation and credential security, rather than a demo-only chatbot.

02

Constraints

  • The assistant had to perform real operations, so read-only question answering was not enough.
  • Multi-tenant isolation and credential security had to be designed before any AI tooling was built on top.
  • The platform had to hold up in production, rather than as a demo.

03

Scope

  • A multi-tenant SaaS architecture with role-based access control across five permission levels.
  • Enterprise-grade credential handling: AES-256-GCM encryption and API key management.
  • Specialized Claude AI tools covering Odoo CRUD operations, workflow automation, analytics, and batch operations, exposed through a custom tool-calling system.
  • A streaming chat interface over Server-Sent Events with reconnection handling.
  • A Gateway API for service-to-service integration with rate limiting.
  • Unit, integration, and end-to-end test suites, then production deployment on Render with CI/CD via GitHub Actions.

04

Approach

Stack

Next.js 15 monorepo 路 seven modular packages 路 Claude tool calling 路 SSE streaming 路 Render

  1. 01

    Architecture & Security Design

    Designed a multi-tenant SaaS architecture with role-based access control (five permission levels) and enterprise-grade credential handling with AES-256-GCM encryption and API key management, before any AI tooling was built on top of it.

  2. 02

    AI Tool Development

    Built more than twenty-five specialized Claude AI tools covering Odoo CRUD operations, workflow automation, analytics, and batch operations, exposed through a custom tool-calling system.

  3. 03

    Real-Time Interface & Gateway API

    Implemented a streaming chat interface over Server-Sent Events with reconnection handling, plus a Gateway API for service-to-service integration with rate limiting.

  4. 04

    Testing & Deployment

    Reached coverage above eighty percent across unit, integration, and E2E tests, then deployed to production on Render with auto-scaling and CI/CD via GitHub Actions.

05

Evidence

A working production platform: a Next.js 15 monorepo of more than fifteen thousand lines across seven modular packages, more than fifty REST API endpoints, and more than twenty-five AI-powered Odoo tools.

A full report-generation pipeline covers PDF, Excel, CSV, and PNG output, and coverage sits above eighty percent across the unit, integration, and end-to-end suites.

06

Limitations

  • Internal platform build: there was no client engagement to oversee, and no client outcome metrics exist.
  • This study demonstrates the custom-software-around-the-ERP capability that goes into client integrations and bespoke modules, rather than an Odoo client deliverable.

Start a project

Have a project like this one?

Tell us which document your team retypes into your ERP. A senior engineer replies with a written plan. Fixed price, agreed before we start.

Start a project

Free written plan and a quote. No meeting.

Back to all case studies