Quickstart: First Agent in 10 Minutes

This guide takes you from a blank console to a running AI agent. By the end, you’ll have created an agent, given it instructions, and run your first task.

Prerequisites: An Auteryn account. Sign up with Google; we review every new team by hand, usually within a day.

Prefer a template?

This guide builds an Internal agent from scratch. If you’d rather start from a template and deploy a customer-support agent to your website, follow the Interactive Quickstart Tutorial →.

---

Step-by-step

1. Create a new agent

   Sign in to the Auteryn Console. If you have no agents yet, the home screen asks What should your first agent do? For this guide, choose Start blank under the question. (Later, choose New agent on the Agents page and find Start blank at the end.)

   Give the agent a name, like “My research assistant”, and answer What should this agent do? in one line. Under Who will talk to it?, keep Your team (an Internal agent). Choose Create and try it. The agent’s first chat opens in the same tab, in the chat workspace.

   Agent types

   Internal agents are for your team’s use — with an optional switch that lets people outside your team join specific conversations. Customer-facing agents are for deployment to external users via channels. You can’t change the type after you create the agent.

2. Write instructions

   Go back to the Console, open Agents and choose your agent. Paste this example into its Instructions section:

   You are a research assistant. Your job is to:
   - Answer questions clearly and concisely
   - Search the web when you need up-to-date information
   - Summarize long documents into bullet points
   - Always cite your sources
   
   Be professional but conversational. If you're unsure about something, say so.

   In its Capabilities section, turn on Web search. Choose Save changes.

3. Run your first task

   Go back to the chat: Chat at the top of the agent’s page, or Open your chats at the bottom of the sidebar.

   In the chat:

   • Type a question: What are the most popular open-source AI agent frameworks right now?
   • Press Enter
   • Watch the agent think and respond in real time

   The agent’s computer may still be starting on its first run, so the first reply can take a little longer. You’ll see its steps, like web searches, appear inline as they happen.

4. (Optional) Add Knowledge

   If you want the agent to answer questions about your specific content:

   • Click Knowledge in the sidebar (under Build), then New collection
   • Upload a PDF or paste a URL to crawl
   • Go back to your agent → Knowledge section → attach the collection
   • Ask a question specific to your document

5. (Optional) Connect an integration

   To give your agent access to external services:

   • Click Integrations in the sidebar (under Connect)
   • Open a service such as GitHub, Slack or Google Drive, and choose Connect
   • Follow the sign-in window
   • Go back to your agent → Integrations section → enable the connection

---

What just happened?

When you ran the task, the agent:

1. Read your instructions to understand its role
2. Classified the task complexity and selected the right execution mode
3. Searched the web for up-to-date information (if web search is enabled)
4. Synthesized the findings into a response
5. Streamed the response back to you in real time

---

What’s next?

Better Instructions

Learn how to write instructions that make your agent far more reliable. Writing Instructions →

Connect Integrations

GitHub, Slack, Google, Jira — give your agent real capabilities. Integrations →

Deploy It

Publish your agent to Telegram, WhatsApp, or a web widget. Internal agents can use Slack via Channels. Deploy →

Automate

Run your agent on a schedule, or when something happens in a connected service such as Slack. Loops →