# Get started for builders

The path from opening Studio to a published agent that answers from your organisation's knowledge and calls a tool, in the order a builder actually does it.

This page walks a builder from the first look at Studio to an agent that Members can pick in the Work Room. It is a journey, not a reference: each stage builds on the one before, links to the fuller page, and ends with something you can check.

**Who can use it:** the **Builder** role or higher, in a tenant licensed for agents and knowledge. You also need to belong to a team, because every agent needs an **Owner team** before it can be published; if you are in no team, ask your administrator to add you to one in the Control Room.

Video: Get Started for Builders (2:46), https://katonicdocsmedia.blob.core.windows.net/docs/get-started-for-builders/6ac62aaf0683/index.html

## Before you start

- Check your role. **Switch workspace** (the grid icon) in the top bar lists **Studio** for a Builder. If it offers only the Work Room and the Catalog, you are a Member; ask your administrator for the Builder role.
- Knowledge needs an embedding model, which your administrator configures under the Settings tab of Knowledge Management. Until one is set, the source wizard cannot be used.
- Have some content to answer from: a few files to upload, or the credentials for a connector.

![The Studio Dashboard with the role chip reading Builder and the Switch workspace menu open: three zone cards, Work Room, Studio marked Current and Catalog.](https://docs.katonic.ai/img/get-started/builders/builder-zone-menu.png)
*A Builder's Switch workspace menu lists Studio alongside the Work Room and the Catalog.*

## The stages

  
**Step 1. Open Studio and read the Dashboard**

Switch to Studio. The **Dashboard** shows what needs a builder's attention, the week's figures, the fleet of agents and the drafts you last edited; **Or start something new** offers **New agent**, **Run an eval**, **Open Prompt Studio** and **Add knowledge**. **My Agents** lists every agent in the workspace with its status. [The Studio Dashboard](/agents/studio-dashboard) and [My Agents](/agents/my-agents).

    ![Studio's Dashboard as a Builder: the sidebar with Dashboard, My Agents and the Develop, Deploy, Improve and Resources groups, the Builder chip in the top bar, the greeting line with the workspace summary, and the What needs you section.](https://docs.katonic.ai/img/get-started/builders/builder-studio-dashboard.png)
*Studio as a Builder sees it: the Dashboard's summary line and the items that need attention, with the Builder chip in the top bar.*

  
**Step 2. Add your first knowledge source**

Open **Knowledge Sources** and select **Add Source**. The wizard has four steps: **Select Source**, **Configure**, **Schedule & Access** and **Review**. Start with **File Upload**, name the source, add a few documents, choose a one-time sync and keep the source **Personal** until it works. Check: the source's row reads **Synced** with a document count. [Add a knowledge source](/knowledge/knowledge-bases/add-a-knowledge-source); the connectors have [a page each](/knowledge/connectors/overview), and [Control who can use a knowledge source](/knowledge/knowledge-bases/access) covers sharing it with teams later.

  
**Step 3. Create the agent**

Select **New agent**. Pick **Answer & assist**, then **Search your docs** for an agent that answers from documents, or **Answer questions** for a general assistant. Choose a template or the **Blank** card and select **Continue to build**. Check: the builder opens on your draft, and My Agents lists it as **Draft**. [Create an agent](/agents/create-an-agent).

  
**Step 4. Give it an identity, a prompt and your knowledge**

In **Identity**, set the **Display name**, the slug, a one-sentence **Description** and the **Owner team**. In **Behavior**, write the **System prompt** or select **Generate from Identity & template** for a starter, and leave the **Model tier** on **Balanced**. In **Knowledge**, switch your new source on under **Pick knowledge bases**. Check: the **Agent draft** rail on the right lists the source. [Build an answer agent](/agents/build-an-answer-agent).

  
**Step 5. Attach a tool**

In **Tools**, three groups are offered. **Connected servers** lists the MCP servers an administrator has registered in the Control Room; pick the tools this agent may call, and note that write actions get an approval gate by default. **Function tools** are the built-in tools that run inside Katonic, such as **Document Generator**. **Agents** attaches published agents this one can call. If the page says **No MCP servers connected yet**, ask an administrator to register one; the Catalog's **Tools** section shows which servers are on offer and whether each runs here. [Tools and MCP](/tools/overview) and [Tools and connectors in the Catalog](/user-guide/catalog/catalog-tools-and-connectors).

  
**Step 6. Test it in the sandbox**

Open the **Test** tab, type a question your documents can answer and select **Run**. The answer streams in and the **Run trace** shows the configuration, the tool calls and the model call of the real run. Save the prompts you want to rerun. Then open **Deploy** and read **Pre-flight**: it names what is still missing, such as a guardrail profile, which you pick in the wizard's **Guardrails** section from the profiles an administrator created. Check: a run in the Test tab answers from your source, and Pre-flight lists no blockers. [Test, publish and versions](/agents/test-publish-and-versions).

  
**Step 7. Publish**

Select **Publish** in the build header. Under **Who can use it**, choose **Team** or **Organization**; under **Where it runs**, tick **Work Room**, **REST API** or both. Fix any blockers the dialog lists, then select **Publish to production** or **Publish to** the lane your tenant uses. The success screen says whether the version **is now live**, was published to a lane to **promote when ready**, or is **not live yet** and waiting on approval. Check: **Open in Work Room** starts a conversation with it, and it appears on the Agents page for the people you chose. [Test, publish and versions](/agents/test-publish-and-versions) and [Using agents](/user-guide/work-room/using-agents).

  
**Step 8. Keep it good**

Run an evaluation against a dataset of questions and expected answers, with a pass threshold, so a change to the prompt or the model shows up as a number. Read the agent's page for its traces and versions, and roll back if a version misbehaves. [Evaluations](/agents/evaluations), [The agent's page](/agents/agent-page) and [Profiler](/agents/profiler).

Each stage ends with a check you can make yourself, and each links to the fuller pages in the Agents, Knowledge and Tools sections.

## Related

- [Agents](/agents/overview)
- Models & Compute
- [Knowledge](/knowledge/overview)
- [Tools and MCP](/tools/overview)
- [Get started for administrators](/get-started/overview)

Source: https://docs.katonic.ai/get-started/builders
