# Agents

To create a custom agent, head to the navigation tab **Agents →** click **Create agent.** You can also see this [video walk-through](#video-walk-through) with an example use case. 

{% stepper %}
{% step %}

### Create an agent

It'll be easier to navigate if you give your agent a proper name and a description.\
\&#xNAN;*Note: you can create as many agents as you'd like: different tasks, use cases, etc.* 
{% endstep %}

{% step %}

### Execution

Choose how to trigger your agent. Currently, we support **prompt-based triggers**, which let you chat with the agent or run now.

**What does “Run Now” mean?**\
Use it when the agent can follow instructions immediately.\
\&#xNAN;*Example:* A Sprint Architect agent that checks JIRA and summarizes sprint health can run now automatically.
{% endstep %}

{% step %}

### Model

Choose the model your agent will use to process prompts. You can change it at any time.\
Model fallback behavior follows your organization’s settings (i.e., if one model fails, another will be used).

:bulb:*Pro tip*:  *if you want to get the best outcomes from agents, use the models our AI engineers recommend. You can access the recommended models from the model picker in the agent creation form.*
{% endstep %}

{% step %}

### Creativity

AI models generate text by predicting the next word based on probabilities. You can control how strictly the model follows your **context** and how much it draws on **outside knowledge**, which affects the novelty and length of the output.

*Grounded*: strictly context, concise and safe; *Guided*: mostly context with selective knowledge; *Balanced*: mix of context and outside knowledge; *Analytical*: context-focused, uncovers patterns; *Creative*: freely explores new ideas.
{% endstep %}

{% step %}

### Instructions

Describe how your assistant should follow prompts, generate responses, and maintain tone. You can enter up to 32,000 characters to set style, rules, and behavior. Note, that longer instructions tend to generate longer response.

:bulb:*Pro tip*:  LLMs understand Markdown — use headings, lists, and formatting to clearly communicate your instructions.
{% endstep %}

{% step %}

### Capabilities

If needed, enable web search to include real-time results from the internet.\
Enrich your agent with knowledge base - add files from an existing [Project](/workspace/projects.md), like brand guidelines, tone of voice, or reference documents to give your assistant extra context.\
This lets it provide sourced answers using your own materials. Knowledge can be updated or removed anytime.
{% endstep %}

{% step %}

### Integrations

Empower your agent with [integrations](/gateway-api/integrations.md) and allow connect to popular platforms like Microsoft 365, SharePoint, Jira, Confluence, Slack and many more. Your agent can fetch information, view files, and read messages. These integrations let your agent provide context-aware answers and insights from your connected systems.\
Note: if you want to use integrations, select a model that supports them.
{% endstep %}
{% endstepper %}

Once saved, you can start chatting with your agent right away. It will appear in the agents list under the Agents menu, and you can access all your agents via the **Chat** model dropdown.

### **Sharing** 

You can share Agents with your team or users.

* View access – others can use the Agents but can’t change it
* Edit access – collaborators can update settings, prompts, and sources

This makes it easier to collaborate, standardize workflows, and keep everyone aligned.

### Video walk-through

{% embed url="<https://www.youtube.com/watch?v=uogFpbEhRbA>" %}


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://docs.nexos.ai/workspace/agents.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.