Getting Started

This guide walks you through creating a Kajoo Agentic workspace, connecting your environments, and starting your first AI-assisted delivery workflow.

If you haven't read the Overview yet, start there — it covers what Kajoo Agentic does and who it's for. For a full explanation of platform concepts (workspaces, agents, skills, recipes), see Platform Concepts once you're ready to go deeper.

📘

This guide is written for developers and solution architects setting up the workspace, as well as delivery leads and project managers overseeing the migration. Technical steps are called out clearly — non-technical readers can follow the flow and hand off specific steps to their team.

---

Step 1 — Access Kajoo Agentic

If you're an existing Kajoo Classic subscriber

You already have access. Kajoo Agentic shares the same organization, users, and authentication as Kajoo Classic — no separate account or invitation required.

You'll see a Kajoo Agentic entry in the left navigation of your Classic dashboard, along with an introductory banner. Click either to launch the Agentic workspace. Organization management, team members, and user roles continue to be managed from the Classic interface.

If you're new to Kajoo

Kajoo Agentic is currently available by invitation. Reach out to your Kajoo contact or contact us to get access.

📘

You can switch between Kajoo Classic and Kajoo Agentic at any time. Your organization, users, and billing are shared — there's no separate login or account to maintain.

---

Step 2 — Create a workspace

From the Agentic dashboard, select New Workspace. You'll be prompted to choose how to create it:

Start from a recipe (recommended)

Recipes are pre-configured workspace templates for specific delivery use cases. Choosing a recipe automatically provisions the full agent team, MCP connections, skills, and tool configurations — along with a setup wizard to connect your environments.

For most teams this is the right choice. Everything is pre-configured; you just connect your project and start working.

More recipes are in development. See Recipes for the full roadmap.

Start from scratch

If no recipe covers your use case, you can build a workspace manually — adding your own agents, connections, skills, and prompt templates. Start-from-scratch workspaces are fully supported and can be as simple or sophisticated as your workflow requires.

📘

You can use an existing recipe workspace as a reference when building from scratch.

---

Good to know before you start

Sessions are resumable. Every phase saves progress as artifacts in your workspace. Open a new chat session at any time and agents pick up from where you left off.

Human-in-the-loop is configurable per tool. High-impact tools require your approval before executing by default. Routine reads are auto-approved. Adjust any tool's setting in the agent configuration panel.

Multiple workspaces and projects are supported. Create one workspace per use case and one project per engagement or client site. Workspaces are fully isolated from each other.

Don't run two code migration sessions in parallel. Code migration sessions share the same sandbox — running them concurrently causes conflicts. You can run code migration and content migration in parallel, as they operate on independent systems.

---

Step 3 — Connect your environments

When you create a workspace from a recipe, the setup wizard guides you through connecting the environments that recipe needs — typically source and target code repositories and CMS instances.

Each recipe has its own environment setup guide with the exact steps and credentials required:

📘

You don't need environments fully ready before creating a workspace. Create it, explore the configuration, and connect environments when they're ready.

---

Step 4 — Explore your workspace

Once your workspace is created you'll see three main areas:

Agents — the AI workers provisioned for this workspace. Each agent has a role, tools, skills, and an LLM assigned to it. If you started from a recipe, everything is pre-configured. You can review, adjust, or extend any agent at any time.

Connections — the MCP servers registered in this workspace. Each connection exposes tools that agents can call. You can add custom connections here — for example, an Atlassian MCP server to link Jira tickets to delivery work.

Projects — the working areas inside the workspace. A workspace can have multiple projects (one per client site or engagement), and each project has its own chat sessions. Create a project, open a chat session inside it, and you're ready to work.

For a full explanation of each concept — including workspaces, agents, skills, recipes, and MCP connections — see Platform Concepts.

---

Step 5 — Start working with agents

From a chat session inside a project, interact with agents by typing naturally. Each recipe has a designated starting point — a start command that invokes the recipe's orchestrator and kicks off the workflow.

The orchestrator will:

1. Check that your workspace is correctly configured
2. Ask for any missing configuration
3. Begin the first phase of the workflow
4. Guide you through each step, asking for your approval before any meaningful action

You don't need to know which sub-agents are involved or what tools they're calling. The orchestrator manages that. Your job is to review what it produces and decide when to proceed.

---

What's next