Documentation

Automations that run on their own — with governance

NNumbers Agents is where you build automations combining AI and deterministic steps, publish them as versions, schedule them, expose them over an API, and follow every run. All with the same identity and the same policies as Imaginne.

Build your first agentWhat Agents is

This is the official documentation for people who build, publish, and operate agents. It covers everything from your first agent to the error reference and the space's governance.

Choose where to start

💬Build by conversation

Describe what you want and Imaginne builds, tests, publishes and activates it.

Learn how →🚀Quickstart

Use a ready-made template, test it, and activate — in a few minutes.

Start →🧭Get to know Studio

Conversation, simple mode, the visual editor, and the Save · Test · Activate cycle.

Explore →🧩Step types

The thirteen available steps and when to use each.

See the steps →

An agent's path

🛠️Build

Steps, data between steps, branching, approvals, and error handling.

Open →📦Publish

Immutable versions, separate environments, and promotion with evidence.

Open →⏱️Trigger

Manually, on a schedule, over an HTTP endpoint, or called by another agent.

Open →📊Operate

Follow runs, decide approvals, cancel, and retry.

Open →

Governance and reference

🛡️Spaces and access

Agent Spaces, roles, and who can build, publish, and run.

Open →🔌Connections

Server and credential under a name of their own — the password goes in once and does not come back.

Open →🔗API

Trigger agents over HTTP and follow runs from an integration.

Open →🩺Execution errors

What each code means and what to do.

Open →

The core idea

An agent is a versioned definition of behavior — what it does, with which data, within which limits, and under which identity. Scheduling it, exposing it over an API, or calling it from another agent are ways of triggering the same definition, not different products.