Agents

Every setting explained

The seven steps of the creation wizard, one by one and on video: identity, mission and model, trigger, permissions, limits, notifications, review.

4 min readUpdated November 13, 2026

The creation wizard has seven steps. Whatever path you choose (template, plain-language description, blank form or import), you go through the same ones, prefilled or not. Everything can still be changed after creation, from the agent's page.

Identity

The agent's name, its avatar and above all the Agent's Odoo account. The agent borrows the rights of this account, and its actions appear under that name in the Odoo history. No extra license is needed, but this account sets the agent's scope: choose it according to what the agent should be able to touch.

Identity step: rename the agent and choose the Odoo account whose rights it borrows.

Only accounts whose member has saved their personal Odoo token in Aidoo can be selected, and only owners and admins make this choice.

Role & AI model

The mission tells the agent what to do, in plain language: what to look for, what to do with each item, how to conclude. It is the setting that matters most. It accepts up to 20,000 characters.

Role & AI model step: complete the mission, then keep the Auto model.

The Auto model suits most missions: Aidoo picks a cost-effective model suited to agents and keeps it up to date. Choose a specific model only if the results remain approximate. Writing tips and model choice are detailed in Creating an agent.

Trigger

When the agent works:

  • Scheduled: a ready-made rhythm ("Every weekday morning (8am)", "Every hour"…) or Custom, up to 5 slots.
  • Event-driven: when an Odoo record is created or updated, with optional conditions, or at a date of the record.
  • Conversational: the agent only answers when spoken to, from Aidoo or from the Odoo bubble.
Trigger step: a ready-made or custom schedule, an Odoo event, or conversation only.

An event-driven agent reacts with a delay that depends on your tier: one hour on Starter, fifteen minutes on Pro, five minutes on Business. See Credits and consumption.

Permissions

The Learning mode box is checked by default. For 48 hours, the agent reads everything its Odoo account sees, and every write that no rule covers is submitted to you. At the end of the window, what it actually did becomes its rules. The rules of a gallery template stay in place as starting rules.

Permissions step: learning mode, the template's starting rules and human approval for writes.

The same step offers optional capabilities, all off by default: Human approval for writes (every write waits for your approval), Web access, Suggest emails (approved sending), Document reading, questions to a human and Aidoo Drive. Details are in Rules and permissions.

Limits & safety

The safeguards of a run: Max credits / run (100 by default), Max duration (seconds) (300 by default), Max writes / run and number of actions. An Agent monthly budget (credits) comes on top of the workspace pool, and the Automatic pause stops the agent after several failures in a row (3 by default).

Limits & safety step: a monthly budget for the agent and the automatic pause after repeated failures.

When starting out, a cap of 5 writes per run lets you see what the agent changes before giving it more leeway.

Notifications & access

Where the agent reports: email, notes in the Odoo chatter, a Discuss channel. The same step picks the Odoo users who will see the agent in the Aidoo bubble of their Odoo; Select all checks the whole team.

Notifications & access step: where the agent reports, and who sees it in Odoo.

Review

A last read-through, then Create the agent: it is created and activated right away, with its own API key, whose permissions follow its rules exactly. If activation is refused (no Agents subscription, insufficient tier, active agent limit reached), the agent stays a draft and its page shows the reason.

The review, then Create the agent: it is active right away.

Test before the first slot

No test is launched automatically. From the agent's page, Test (dry run) runs the mission while simulating every write. See Your first agent.

Next step