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

# Connectors

> See which tools the AI Studio agent can use, install agent-enabled connectors, and add external MCP servers

The **Connectors** page in AI Studio lists the security tools the agent can run in this workspace. Open it from **Connectors** in the AI Studio sidebar; the page header reads "Tools the agent can use in this workspace".

![AI Studio Connectors page with tool cards for arjun, dirsearch, ffuf, gau and katana (each with a Config button and a menu arrow), gitleaks, nikto and nmap (each with an Install button), and httpx and naabu (each marked Built-in), with category badges such as Url Discovery, Vulnerabilities, Http Probe and Ports Scanner](https://mintcdn.com/oasmplatform/JcAER2Ib37AN8_8A/images/screens/ai-connectors.webp?fit=max\&auto=format\&n=JcAER2Ib37AN8_8A\&q=85\&s=4c1eba60df66a6ca82dc21368a906ed9)

The Connectors page: installed connectors show Config, built-in tools show Built-in, and available connectors show Install

## What the page lists

| Kind | Which ones appear | Button on the card |
| - | - | - |
| **Built-in tools** | Every tool built into the worker. Always available, cannot be removed | **Built-in** (disabled) |
| **Agent connectors** | [Connectors](https://docs.oasm.io/tools.md) whose manifest marks them as usable by the agent, installed or not | **Install** when not installed; **Config** with a menu arrow when installed |

Connectors that are not enabled for the agent do not appear here; you still find them on the [Tools](https://docs.oasm.io/tools.md) page for use in [projects](https://docs.oasm.io/project.md) and [jobs](https://docs.oasm.io/jobs.md).

Each card shows the tool's name, logo, a check mark for officially supported tools, its description and its category badges. Clicking the card opens the tool's detail page.

## How the agent uses these tools

* The agent can run **built-in tools** and **installed** agent connectors. It refuses a connector that is not installed in the workspace.
* Running a tool needs **Agent mode** and an online [worker](https://docs.oasm.io/worker.md); in Ask mode the agent only reads workspace data. See [Chat](https://docs.oasm.io/ai-studio/chat.md).
* When the agent runs a connector, it picks a configuration profile or fills in non-secret settings itself. A run is refused while a required setting is still empty, so configure secrets such as API keys in a profile first.
* Every tool call goes through the same safety checks and, in Manual mode, the same approval prompt. See [Plans and approvals](https://docs.oasm.io/ai-studio/plans-and-approvals.md).

## Install a connector for the agent

**1. Open Connectors**

In the AI Studio sidebar, click **Connectors**.

**2. Install**

Click **Install** on the connector's card. The **Install \<name>** dialog shows the workspace, what the tool does and the data shared with it. Click **Install** to confirm. The console confirms with "Tool installed successfully" and you stay on the page.

**3. Configure it**

Click **Config** on the card to open the configuration sheet and create a configuration profile, for example with the tool's API key. Connectors that need no settings can be used right away.

To remove a connector, click the menu arrow next to **Config**, select **Uninstall**, and confirm in the **Remove Tool** dialog. This removes it from the whole workspace, not only from the agent.

| Action | Requirement |
| - | - |
| Open the Connectors page | `agent.read` |
| Install or uninstall a tool | `workspace.write` |

See [Permissions](https://docs.oasm.io/permissions.md).

## Connectors and the Tools page

The Connectors page is a filtered view of the same workspace tool catalog as **Tools**. Installing, configuring or uninstalling a tool here is the same action as on the [Tools](https://docs.oasm.io/tools.md) page, and the change applies to scans and projects as well as to the agent. The one difference: installing from **Tools** takes you straight to the new connector's configuration, while installing from AI Studio keeps you on the list.

To package your own tool as a connector, see [Connectors overview](https://docs.oasm.io/connectors/overview.md).

## External MCP servers

The agent can also use tools from external servers that speak the Model Context Protocol (MCP). These servers are configured per workspace, and every member's main agent gets their tools.

There is no console screen for this yet. Manage the servers through the API under `/api/agents/mcp-configs`:

| Method and path | What it does | Permission |
| - | - | - |
| `GET /api/agents/mcp-configs` | List the workspace's MCP servers | `agent.read` |
| `PUT /api/agents/mcp-configs/{name}` | Add or update the server called `name` | `agent.write` |
| `PATCH /api/agents/mcp-configs/{name}/toggle` | Turn a server on or off (`{"disabled": true}`) | `agent.write` |
| `GET /api/agents/mcp-configs/{name}/ping` | Check whether the server is reachable (`online`, `offline` or `unknown`) | `agent.read` |
| `DELETE /api/agents/mcp-configs/{name}` | Remove the server | `agent.write` |

The body of `PUT` accepts these fields:

| Field | Description | Default |
| - | - | - |
| `url` | The server's HTTP or HTTPS endpoint | |
| `transport` | `sse` or `streamable-http` | `sse` |
| `headers` | HTTP headers sent with every request, for example an API key header. Values must be strings | |
| `disabled` | Keep the server configured but unused | `false` |
| `allowed_tools` | Only expose these tool names. Leave unset to expose every tool | all tools |
| `timeout` | Connection and request timeout, in seconds | `60` |
| `sse_read_timeout` | Maximum wait for an SSE event, in seconds | `300` |

```json theme={null}
{
  "url": "https://mcp.example.com/sse",
  "transport": "sse",
  "headers": { "Authorization": "Bearer <token>" },
  "allowed_tools": ["lookup_host"]
}
```

How external tools behave in a chat:

* Each tool is named `<server>_<tool>`, for example `intel_lookup_host` for the tool `lookup_host` on a server named `intel`. If that name collides with a built-in tool, the built-in tool wins.
* External tools go through the built-in deny list, the workspace tool rules (managed by the workspace owner through `/api/agents/tool-rules`) and, in Manual mode, the approval prompt, which shows **Allow \<tool>?** with the arguments and the server the tool comes from.
* Only the main agent gets external MCP tools; subagents do not.

## Using OASM from your own AI client

The servers above are tools the AI Studio agent *calls*. The opposite direction, letting an external AI client such as Claude or Cursor read your OASM workspace, is the OASM MCP server. See [MCP Server](https://docs.oasm.io/settings/mcp.md).

## Related

> [card] [Tools](https://docs.oasm.io/tools.md)
>
> The full tool catalog, configuration profiles and tool details

> [card] [MCP Server](https://docs.oasm.io/settings/mcp.md)
>
> Connect external AI clients to your workspace

> [card] [Workers](https://docs.oasm.io/worker.md)
>
> Where the agent's tools run

> [card] [Plans and approvals](https://docs.oasm.io/ai-studio/plans-and-approvals.md)
>
> Approval prompts for commands and tool calls


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