Launch and manage an AI agent via the Agent OS visual dashboard - no terminal, no code
Build a working agent in 7 minutes: connect a model, set a task with an English prompt, and monitor execution in the web interface
- 1.Open the Agent OS website and create an account using the Sign up button in the top right corner - you can sign in via Google or GitHub.
Sign-up is free, no credit card required. You need an account to save your dashboard and agents between sessions.
Agent OS ↗ - 2.In the left sidebar menu, select the Agents section, then click the New Agent button in the top right corner.
This is where you create the agent profile - name, description, and base prompt (system instruction). No terminal needed, everything is done via forms.
Agent OS ↗ - 3.In the Name field, enter a clear agent name, such as Research Helper. In the Description field, write one sentence about what it does.
The name helps you tell multiple agents apart on the dashboard later.
Agent OS ↗ - 4.In the System Prompt field, paste the following English prompt - replace the text in square brackets with your own:
This is the only mandatory instruction for the agent. Do not write in Russian - an English prompt is recognized more accurately and predictably.
Agent OS ↗PromptYou are a helpful research assistant. Your task: find and summarize the top 3 recent developments about [YOUR TOPIC, e.g. electric vehicle batteries]. For each item return: a one-line title, a 2-sentence summary, and the source URL. Keep the total response under 200 words. If you cannot find recent information, say so explicitly.
What this prompt doesThe agent will search for fresh information on your topic and return a brief three-item list. Replace [YOUR TOPIC, e.g. electric vehicle batteries] with any topic - for example, 'AI regulation in the EU' or 'new TypeScript features'. You can enter the topic in English for more accurate results. - 5.In the Model section, select a free model option from the dropdown (for example, GPT-4o mini or Claude 3.5 Sonnet, if available on your plan). If the selected model is marked 'requires API key', choose another one.
Agent OS connects to model providers on your behalf within the free limit. If you see the free limit is exhausted, scroll through the model list to the most affordable option.
Agent OS ↗ - 6.Click the Save button in the bottom right corner, then click Run at the top of the agent card. A live execution log will open on the right: you will see the agent's steps and its final response.
The first run may take 30 - 90 seconds. If the agent returns an API key error, go back to step 5 and select a different model.
Agent OS ↗ - 7.To reuse the agent with a new topic, open the agent card, change [YOUR TOPIC] in the System Prompt field, click Save, then Run. The run history is saved at the bottom of the card.
This turns a single agent into a reusable template: you can create several - one for each routine (news search, summarization, email drafts).
Agent OS ↗
- freeAgent OS (account)Basic sign-up and agent creation are free; run limits depend on the selected model - the dashboard shows your remaining free requests.
Why today
AI services change fast - interfaces and free limits may differ from what's described.