> ## Documentation Index
> Fetch the complete documentation index at: https://docs.enclave.ai/llms.txt
> Use this file to discover all available pages before exploring further.

> ## Agent Instructions
> Enclave in these docs is the code and cloud security platform at enclave.ai. It is not Open Enclave, Intel SGX or other hardware enclaves, or enclave.gg.
> People use Enclave in the web app at https://app.enclave.ai or the desktop app. These docs describe no public Enclave API or CLI, so don't invent endpoints, commands, or config files.
> Connecting an integration needs a person to sign in and approve access in a browser. Tell the user the exact step to take instead of saying you completed it.

# Agents

> Built-in and custom agents: their models, prompts, and tools.

An agent is a model with instructions and a set of tools. Agents do the work inside [workflows](https://docs.enclave.ai/workflows/overview.md). Click **Agents** in the sidebar to see your library.

![The Create agent form with details, instructions, and tools](https://mintcdn.com/enclave/bhnFMg4ldiTYrwZX/images/product/create-agent.png?fit=max\&auto=format\&n=bhnFMg4ldiTYrwZX\&q=85\&s=57128226bc1158d7b980fc57542ebaa2)

## Built-in and custom agents

* **Built-in agents** power Research, Slack, PR review, repository learning, and security hunts. You can change their model and reasoning level for your organization. Click **Reset to default** to undo.
* **Custom agents** are yours. You control everything, including the prompt and tools.

## Create an agent

**1. Open the form**

Click **Create agent**.

**2. Describe the agent**

Enter a **Name**, an optional **Description**, and pick a **Model** and **Reasoning level**.

**3. Write the instructions**

In **System prompt**, define the agent's role, focus, and how it should respond.

**4. Pick tools and save**

Select the tools it needs, then click **Create agent**.

You can also keep an agent from a workflow proposal. Click **Save to agent library** on its **Agents** tab.

## Models

The default is Sonnet 5.5.

| Provider | Models |
| - | - |
| Anthropic | Opus 5.5, Sonnet 5.5, Haiku 5.5, Fable 5 |
| OpenAI | GPT-5.6 Terra, GPT-6 Astra, GPT-6 Luna, GPT-6.1 Sol |
| Moonshot AI | Kimi K3 |
| Z.ai | GLM 5.3 |
| DeepSeek | DeepSeek V4.1 Flash |

**Reasoning level** sets how much the model thinks before it answers: **Default**, **Low**, **Medium**, **High**, **Extra high**, or **Max**. Higher is slower and costs more. It isn't available for Haiku 5.5 or the Kimi, GLM, and DeepSeek models.

## Tools

| Group | What it's for |
| - | - |
| **Code & repositories** | Read files, search code, and run shell commands in a sandbox with the repository checked out |
| **Findings & intelligence** | Read findings and CVE intelligence |
| **Cloud & architecture** | Query the graph of your cloud resources and their exposure |
| **Knowledge & memory** | Search saved memories and read workflow records |

Use the search box to find a tool, or check **Show tool IDs** to see each tool's ID.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.