> ## Documentation Index
> Fetch the complete documentation index at: https://agenvik.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# Build your first AI agent — Agenvik quick start
> Go from sign-up to a live Agenvik agent in under 30 minutes. Upload knowledge, train, tune behaviour, and embed on your website.
This guide takes you through the first setup flow: create an Agent, add Knowledge Sources, train it, tune Behaviour, and prepare a Channel.
## What you will learn
* How to create a new Agent from the dashboard
* How to add Text and Q\&A Knowledge Sources
* How to start training and test the result
* How to prepare the Web Channel for launch
## Before you start
Sign in to the Agenvik dashboard and open **Agents** from the sidebar. If your workspace has reached its active agent limit, archive an unused agent or review your subscription before creating another one.
## Step 1: Create an Agent
1. Open **Agents**.
2. Click the create button in the page header.
3. Enter a clear agent name, such as `Sales Support Agent` or `Product FAQ Agent`.
4. Save the Agent.
Use one Agent per distinct use case. This keeps knowledge, behaviour, and analytics easier to understand later.{/* LEAD_GENERATION_DISABLED: original text included "leads," — restore when re-enabling */}
## Step 2: Add Knowledge Sources
Open the Agent and find **Knowledge Sources**. Agenvik supports:
* **Text**: free-form content, policies, instructions, product notes, or FAQs written as rich text.
* **Q\&A**: structured question-and-answer pairs for exact responses and common customer questions.
* **PDF**: upload a PDF document and Agenvik indexes the full content automatically.
* **Website URL**: paste a URL and Agenvik reads and indexes the page (Starter and above).
Click **Add**, choose the source type, and follow the prompts to save the content.
## Step 3: Train the Agent
After at least one Knowledge Source is ready, click **Train**. Agenvik queues ready sources and indexes them for the Agent.
Training uses your plan's daily training limit. If the button is disabled, check whether the Agent is locked, whether any sources are ready, or whether today's training runs are already used.
## Step 4: Tune Behaviour
Use **Behaviour** to define how the Agent should answer.
* **System Prompt**: the Agent's role and core instructions
* **Tone**: friendly, professional, or neutral
* **Verbosity**: short, balanced, or detailed
* **Fallback Message**: what the Agent says when it cannot answer
* **Guardrails**: boundaries for topics, claims, or escalation rules
## Step 5: Configure the Web Channel
Use **Channel** to control the web chat experience. Set the display name, subtitle, welcome message, input placeholder, and primary color.
The preview on the right side of the Agent page helps you check the greeting, color, and input text before launch.
## Step 6: Review the Agent
Before you share the Agent, ask test questions that reflect real visitors:
* A simple question directly covered by a Knowledge Source
* A question with multiple possible meanings
* A question outside your business scope
{/* - A lead-intent question, if Lead Generation is enabled */}{/* LEAD_GENERATION_DISABLED: uncomment to restore */}
## Next steps
Learn practical ways to improve knowledge, behaviour, and testing.
Add the Agent to your website once it is ready for visitors.