---
title: "Get started"
canonical: https://agentlaire.com/docs/get-started
lastUpdated: 2026-11-03
---

# Get started

## What is AgentLaire?

AgentLaire adds AI agents to your own Odoo database, and every change an agent wants to make waits for a person to approve it. AgentLaire AI Control is the free base module: providers, policies, run records, the audit log and the kill switch. [See the product page](https://agentlaire.com/ai-control).

## What do I need?

AgentLaire AI Control needs Odoo 18, 19 or 20 Community, self-hosted. It cannot be installed on Odoo Online. Enterprise and Odoo.sh support comes later.

| Item | Requirement |
| --- | --- |
| Odoo | 18, 19 or 20 Community, self-hosted |
| Not supported | Odoo Online |
| Model | One provider account or a local model |
| Who sets it up | A user with the AI Manager group |

The administrator who installs AgentLaire AI Control becomes an AI Manager. Nobody else is in an AI group until you add them under **Settings › Users & Companies › Users**.

## How do I install AgentLaire AI Control?

Install AgentLaire AI Control from its Odoo Apps Store listing for your Odoo version. AgentLaire AI Control is free; AgentLaire AI Agents, which adds the agents, chat and approvals, is a paid module ([prices](https://agentlaire.com/pricing)).

1. Download the module from the Odoo Apps Store listing for your Odoo version.
2. Put it in your addons path and restart Odoo.
3. In developer mode, open **Apps › Update Apps List**, then install AgentLaire AI Control.
4. To use agents, install AgentLaire AI Agents the same way.

The new **AI Control** app appears in the main menu.

## How do I connect a model?

AgentLaire AI Control connects to a model under **AI Control › Configuration › Connections › Providers**. Create a provider with these fields:

| Field | What it does |
| --- | --- |
| Provider Type | Anthropic (Claude), OpenAI, OpenAI-compatible (Ollama, vLLM, LiteLLM) or Google Gemini |
| Base URL | The address of the API, for example `http://ollama:11434/v1` for a local model |
| API key | The key from your provider |
| Default Model | The model agents use unless an agent names another |

1. Fill the fields and save.
2. Select **Test connection**.
3. Under Data sharing, select **Allow data sharing**. Until a manager does this, no request with business data goes to the provider.
4. Under **AI Control › Configuration › Settings**, pick the **Default provider**.

For a model on your own server, see [Connect a local model with Ollama](https://agentlaire.com/docs/guides/ollama-local-model).

![Claude provider form in AgentLaire AI Control with budget, health, calls and cost in the header](https://agentlaire.com/_astro/D9.BwKS_I_9_2kkwus.webp)

## How do I set tool policies?

A tool policy tells agents which models and fields they may read or change. Policies are under **AI Control › Configuration › Governance › Policies**. Each policy sets the model, the operations (Read, Create, Write, Delete), the allowed fields and a record limit.

New policies allow no write, no delete and no method calls, and every write is a proposal a person approves. Policies add to Odoo's access rights; an agent never gets more than the user it works for.

With AgentLaire AI Agents, the quickest way is to enable a skill. Open **AI Control › Configuration › Agents › Skills**, open a skill and select **Enable**. This activates the policies of the skill's tools after you confirm the list. The starter policies are created inactive until you do.

## How do I turn on an agent?

AgentLaire AI Agents creates an agent named AgentLaire AI with the starter skills and makes it the default chat agent. Check or change agents under **AI Control › Configuration › Agents › Agents**.

1. Enable at least one skill (see the previous step).
2. Add the people who may use it to the AI User group.
3. Open the agent and select **Test run** to check it.
4. Select **Ask AI** in the top bar and ask for something, for example a quotation.

When the agent wants to change data, it shows a card with Approve and Reject. [Approvals](https://agentlaire.com/docs/approvals) explains how that works.

## What should I read next?

- [Approvals](https://agentlaire.com/docs/approvals): how a proposal works and who can approve it.
- [AI Control](https://agentlaire.com/docs/ai-control): providers, budgets, policies, the audit log and the kill switch.
- [AI Agents](https://agentlaire.com/docs/ai-agents): chat, skills, triggers and the MCP endpoint.
- [FAQ](https://agentlaire.com/docs/faq): buying, versions, data and support.