> ## 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.

# Models and tools

> Choose the model and tools each workflow step's agent uses, and set which tool calls need approval.

Each step in a [workflow](https://docs.enclave.ai/workflows/overview.md) runs an agent: a model with instructions and a set of tools. You set them on the workflow proposal's **Agents** tab when you [create a workflow](https://docs.enclave.ai/workflows/create.md).

## Models

Pick each agent's model and **Reasoning level** on the **Agents** tab. Larger models and higher reasoning levels handle harder tasks, but they're slower and use more credits.

## Tools

| Group | What it's for |
| - | - |
| **Code & repositories** | Read and search code, run commands in a sandbox with the repository checked out, open pull requests, and save reports and files |
| **Findings & intelligence** | Read and triage findings, and read CVE intelligence and the Threat Feed |
| **Cloud & architecture** | Query the graph of your cloud resources and their exposure |
| **Knowledge & memory** | Search the web, and save and search memories |
| **Actions** | File issues in Linear or Jira |

### Approval

A tool that makes changes shows an **Approval** setting once you select it:

* **Don't ask: runs now:** the agent calls the tool right away.
* **Ask later: runs once OK'd:** the call is saved and the agent carries on. The call runs once a person clicks **Approve**, even after the run ends.
* **Ask first: waits for OK:** each call waits until a person clicks **Approve** or **Deny**.

See [Approve tool calls](https://docs.enclave.ai/workflows/runs.md#approve-tool-calls).

> [!WARNING]
>
> Web pages and repository content are written by others and can mislead an agent. Before you give an agent that reads them tools that make changes, set those tools to **Ask first: waits for OK**.

### File issues in Linear or Jira

A workflow can file issues in Linear or Jira where you choose.

**1. Connect the tracker**

Connect [Linear](https://docs.enclave.ai/integrations/linear.md) or [Jira](https://docs.enclave.ai/integrations/jira.md) in **Settings → Integrations**.

**2. Choose where issues go**

On a workflow proposal's **Agents** tab, click **Edit tools** and select the Linear or Jira tool. Pick an account, then the team or project that issues go to.

A saved workflow files each problem once. When a later run finds a problem it already filed, it reuses the existing issue.

### Save reports and files

An agent can save a deliverable, such as a report, a CSV, or a PDF, to the run's **Outputs** tab. When you [create the workflow](https://docs.enclave.ai/workflows/create.md), tell Research what to deliver, for example "save a Markdown report of what you found". See [Results](https://docs.enclave.ai/workflows/runs.md#results).

## Open pull requests

An agent can deliver its changes as a pull request. When you [create the workflow](https://docs.enclave.ai/workflows/create.md), ask Research for it, for example "open a pull request with the fix". The proposal then gives that agent write access to the run's repository.

* Write access works only in a workflow run on a single repository.
* There's no approval step. An agent with write access can push whenever the workflow runs.
* The repository's push rules still apply.

> [!WARNING]
>
> The agent reads repository content and runs the repository's own scripts. Content that misleads the agent can misuse its access. Give write access only to agents that need it.


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