Getting started

Choose the right first build, then create your first agent or workflow in Lunnoa Automate.

Lunnoa Automate helps teams create AI agents and workflows without starting from code. This guide is for a first session: open your workspace, choose a small business process, then build either a first agent or a first workflow.

Choose what to build first

BuildChoose this whenExample
AgentA person should chat, ask questions, or get guided helpA policy Q&A assistant, invoice intake assistant, or supplier support assistant
WorkflowThe same steps should run in order every timeA request approval flow, recurring report, or form-to-notification process

Many real processes use both. A workflow can call an agent, and an agent can use tools that trigger workflow actions. For your first session, choose one starting point so testing stays simple.

First session checklist

  1. Open the right workspace

    Sign in to Lunnoa Automate and confirm the workspace name matches the team or environment you intend to use. If you have access to more than one workspace, use a test or training workspace for your first build.

  2. Pick a project

    Agents and workflows live inside projects. Use an existing project if the process belongs to it, or create a small project for onboarding examples.

  3. Check access

    Make sure you can create the thing you are building. Agents need an available AI Connection and model. Workflows may need app connections before actions can run.

  4. Use demo data first

    Avoid real customer records, production approvals, or sensitive documents while learning. Use a short sample document, a test email, or a simple manual trigger.

Build your first agent

Use this path when the first user experience should be a conversation.

  1. Open Agents

    Go to Agents under Build, then click New Agent.

  2. Choose a project and name

    Select the project and give the agent a job-based name, such as Policy Q&A or Invoice Intake Assistant.

  3. Configure the basics

    Open Configure. Set Profile & instructions, then choose AI settings. The minimum useful setup is a clear name, description, instructions, AI connection, and model.

  4. Add only one useful capability

    If the agent answers from documents, enable one Knowledge notebook. If it should take an action, add one Tool or MCP integration. Do not add everything at once.

  5. Test in Build

    Return to Build and ask three prompts: one easy success case, one case that should use knowledge or a tool, and one case the agent should refuse or escalate.

For the full setup guide, see Create and configure an agent. For every configuration area and option, see Agent configuration reference.

Build your first workflow

Use this path when the first user experience should be a repeatable process.

  1. Open Workflows

    Go to Workflows under Build, then click New Workflow.

  2. Choose a trigger

    Start with Manually Run for your first build. Manual triggers are easier to test because you decide exactly when the workflow runs.

  3. Add one action

    Click the + on the canvas and add one simple action. For example, send a message, create a draft record, or call an agent.

  4. Configure and test the node

    Click the node, fill required fields, connect any required app account, then use Save & Test if the action supports it.

  5. Save, run, and inspect the result

    Click Save, run the workflow manually, then check Monitor executions to confirm what happened.

For a click-by-click canvas guide, see Build on the canvas. If you prefer describing the process in chat, see Chat to workflow.

A good first example

Try one of these:

GoalStart withWhy it is good for learning
Answer a policy questionAgent + one Knowledge notebookShows instructions and document grounding
Collect an internal requestAgent with clear questionsShows conversation design without risky actions
Send a weekly reminderWorkflow with manual or schedule triggerShows triggers, actions, and executions
Ask for approvalWorkflow with Flow Control approvalShows human decision points and run history

Next steps