> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://poly-ai.ferndocs.com/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://poly-ai.ferndocs.com/_mcp/server.

# Tutorial: Create a project

> PolyAcademy Level 1 – Set up your first Agent Studio project with the correct name, language, and voice.

> **Info**
>
> **Lesson 1 of 6** – Create your first project in Agent Studio. This takes about 5 minutes, and the name you choose matters more than you'd think.

> **Lesson**
>
> Level 1 · Beginner · 5 min

![set-up-project](/_fern-files/_img/ecb18dfc9192ac08ef98465f81068a6e2209990c39d61d9b6cec1f82da39be08.webp)

## Create the agent

#### Open the agent creation dialog

Select **+ Agent** to open **Create new agent**.

#### Configure your agent settings

Fill out the fields carefully (these become your default baseline for all later testing):

#### Project name

Use a clear, descriptive name that won't become misleading later.

**Good examples:**

* `Acme Support Agent`
* `Webchat L1 Triage`
* `Hotel Concierge (US)`

**Avoid:**

* `Test`
* `New agent`
* Names that explicitly reference an environment (`Sandbox`, `Prod`, `Live`)

#### Response language

Select the primary language/locale your users will interact in.

#### Agent Voice

Pick a reasonable starter voice. You can always change this later.

#### Greeting

Keep it short and broadly reusable.

* Aim for one sentence that works for both chat and voice
* **Example:** `Hi–how can I help today?`
* Avoid over-specific or wordy greetings

#### Project ID (optional)

Treat this like a stable identifier. It appears in the project URL inside Agent Studio, with an environment suffix appended automatically. If you leave it blank, Agent Studio generates one for you.

> **Warning**
>
> Once set, the **Project ID cannot be changed**. Choose carefully – every bookmarked URL and API call will include this ID forever.

**Allowed characters:** alphanumerics and dashes only (`a-z`, `A-Z`, `0-9`, `-`).

**Examples:**

* `my-shop-support-l1`
* `hotel-concierge-us`

> **Tip**
>
> If you expect multiple similar agents, build a predictable pattern now.

#### Create the agent

Select **Create**.

#### Check your understanding

#### Why should you avoid naming a project 'Test' or 'New agent'?

* The name appears in the live greeting, so a vague name sounds unprofessional to callers
* Generic names cause confusion as you add environments and variants
* The name is permanent and cannot be changed after the first publish
* Agent Studio assigns the Project ID from the name, so a generic name creates a confusing URL

**Answer:** Generic names cause confusion as you add environments and variants

Generic names like 'Test' or 'New agent' become confusing when you have multiple agents in your account. Use a clear, descriptive name from the start – something like 'Hotel Concierge (US)' stays meaningful as you scale.

## Confirm the baseline workspace

After creation, you land in the **Knowledge** area – empty state with an **Add topic** button.

### Agent Studio interface

#### Left sidebar

Navigation for all agent configuration areas (Manage, Build, Channels, Configure, Analytics)

#### Center canvas

Main workspace where you edit topics, functions, flows, and view analytics

#### Right sidebar

Context-sensitive tools like chat testing, call testing, and version history

The right sidebar toggles open/closed for quick access to testing tools.

#### Confirm agent identity

* The header shows the agent name and language.
* The agent is in **Draft** state.

#### Verify navigation access

Use the left navigation to check you can access:

**Behavior**

* **General**: personality, role, and Behavior rules.

**Knowledge**

* Later, you'll learn to create simple FAQ, SMS/handoff topics, and multiturn topics here (using the **FAQs** and **Sources** tabs).
* **Variants**: where you'll configure variants later (even if you ignore it for now, confirm it opens).

**Voice**

* **Agent** (voice/disclaimer selection) and **Advanced > Call settings** (greeting, disclaimer text, and channel-specific settings).

**Deployments**

* Where you'll test and promote changes safely (**Sandbox → Pre-release → Live**, depending on your setup).
* You may also see other areas (Numbers under Voice, Account, Real-time config, Custom Dashboards). You won't need to configure them for Level 1, but you may return to them later.

## Verify

![set-up-project-complete](/_fern-files/_img/27adc49dc0c457022f9db3c76d838e683f96cee99084d8c8d51cd4a03a813bfa.webp)

> **Tip**
>
> **Success criteria:**
>
> * The project loads without errors and stays stable across page changes.
> * You can access (at minimum):
>   * **Behavior > General** (personality, role, Behavior)
>   * **Knowledge**
>   * **Knowledge > Variants**
>   * **Voice > Settings**
>   * **Deployments**
> * The Knowledge page shows an empty state with a clear call to action (for example, **Add topic**), indicating you're ready for the next steps in the checklist.

#### Check your understanding

#### You're naming a new Agent Studio project. Which of these is a bad choice?

* Acme Support Agent
* Linden Hotel Concierge
* Hotel Concierge (US) – Live
* Webchat L1 Triage

**Answer:** Hotel Concierge (US) – Live

Including an environment name like 'Live' in the project name is a problem – the same project runs across Sandbox, Pre-release, and Live. A name tied to one environment becomes misleading in the others.

## Try it yourself

#### Challenge: Name your agent

You're building a hotel concierge agent for "The Linden Hotel". Write out:

1. A project name
2. A project ID
3. A one-sentence greeting

#### Hint

The project name should be descriptive but not reference an environment. The project ID should be stable and URL-friendly (lowercase, hyphens). The greeting should work for both chat and voice.

#### Example solution

**Project name:** `Linden Hotel Concierge`

**Project ID:** `linden-hotel-concierge`

**Greeting:** `Hi, welcome to The Linden Hotel – how can I help you today?`

#### Check your understanding

#### You named your Project ID 'test-1' six months ago. Now your team has multiple agents and everyone bookmarks Agent Studio URLs. What problem will you hit?

* The agent will stop working because 'test-1' is a reserved name
* The Project ID is embedded in the URL – a vague ID like 'test-1' makes it hard to tell which agent a URL points to
* You'll need to re-authenticate every time you open the project
* The ID will conflict with the environment name and block promotions

**Answer:** The Project ID is embedded in the URL – a vague ID like 'test-1' makes it hard to tell which agent a URL points to

The Project ID determines the project's URL inside Agent Studio and cannot be changed once set. A vague ID like 'test-1' becomes confusing when you have multiple agents – every bookmarked URL looks meaningless and you can't rename it later. Choose a descriptive, stable ID from the start (e.g., 'linden-hotel-concierge').

> **Try it**
>
> Project IDs must be URL-friendly. In an ID like "linden-hotel-concierge", words are separated by \_\_\_\_\_.
>
> **Answer:** `hyphens, hyphen, -`

## Go deeper

Want more detail on any of the concepts in this lesson? These reference pages cover the same features in depth:

#### [Agent Studio overview](/agent-studio/get-started/agent-studio)

Capabilities overview and UI walkthrough

#### [Quickstart](/agent-studio/get-started/quickstart)

The shortest path from account creation to a deployed agent

---

#### [← Back to Level 1](/academy/poly-academy/get-started)

Level overview

#### [Next: Edit agent behavior →](/academy/poly-academy/poly-academy-level-1/edit-agent-behavior)

Lesson 2 – give your agent a personality