AI Agent Hub
Back to plugins
⚙️

dsh-odoo

Workflow Updated 2026.08.26

Run the following command in DeepSeek Harness:

dsh plugin install maxmilian/dsh-odoo

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install maxmilian/dsh-odoo in your DeepSeek Harness shell to add the read-only Odoo query plugin; the full source is at https://github.com/maxmilian/dsh-odoo.

About this plugin

Teams that run Odoo increasingly want their AI assistant to pull up contacts, scan orders, or reconcile invoices - but handing write access to an agent feels risky. dsh-odoo exists for exactly that tension: it plugs into Odoo JSON-RPC as a strictly read-only DeepSeek Harness plugin, letting an agent browse contacts, quotations, sales orders, invoices, project tasks, leads, and stock without altering any Odoo state.

Four tools are exposed: read the server version and authenticated user, list queryable fields on an allow-listed model, perform a bounded search_read on that model, and - only when you explicitly opt in via allowWrite - create a tightly constrained draft record (always state=draft, a fixed field subset, no one-to-many commands). Queries are locked to 14 standard models, dotted relational fields and binary fields are rejected, and every response hits a hard byte cap. The safety rails are a capability boundary, not a suggestion.

Built for developers and operators who already run an Odoo instance (verified on 18; versions 8-17 and 19 remain untested) and use DeepSeek Harness. You want your agent to answer read-only questions about last month invoices or project progress, without entrusting it with deletes, workflow transitions, or report generation.

Use Cases

  • Have your AI agent pull up recent invoices, sales orders, and contacts during a conversation
  • Cross-check stock quantities against purchase orders read-only, without logging into the Odoo UI
  • Create a tightly constrained draft project task to help the team log a to-do quickly

Best For

  • Developers who run Odoo and have adopted the DeepSeek Harness
  • Teams that want AI to read ERP data without granting write access
  • Business users who prefer natural-language queries over sales, finance, and project data