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

# Integrations

> Connect apps, skills, plugins, and MCP servers to SuperEmber agents.

Integrations extend what an agent can search, read, and use.

SuperEmber separates three kinds of extensions:

| Type               | What it adds                                                    |
| ------------------ | --------------------------------------------------------------- |
| Connectors         | OAuth-style connected app access and connector triggers.        |
| Skills and plugins | Installable capabilities from the skills marketplace.           |
| MCP servers        | Remote Model Context Protocol servers that expose custom tools. |

## Connectors

Connectors let agents and workflows use external apps.

The dashboard supports searching connector suggestions, opening authorization, checking connection status, reconnecting, disconnecting, and managing connector triggers where a toolkit supports triggers.

Common connector examples include Gmail, GitHub, Slack, Linear, Jira, Notion, Salesforce, and Zendesk. Availability depends on the connector provider and your workspace configuration.

## Connector Triggers

Some connectors can listen for external events.

Triggers are off until you create one. After a trigger is created, incoming events can be sent to the relevant agent, which decides whether the event is useful and what to send back to the user.

When you disconnect a connector, active triggers for that connector are removed or disabled.

## Skills And Plugins

Skills and plugins come from the skills marketplace. Search for capabilities such as MCP, browser, research, memory, files, or app-specific behavior.

Installed items are attached to the agent runtime. After saving, the runtime loads the updated configuration automatically on the next message or after restart.

## MCP Servers

MCP servers expose tools to an agent.

SuperEmber supports remote MCP server configuration when the MCP feature is enabled. Each server has a name, transport/configuration, optional URL, and JSON configuration fields.

## Safety Notes

* Connect only tools the agent is allowed to use.
* Prefer read-only access unless the use case needs writes.
* Treat write actions, HTTP requests, and connector actions as higher risk.
* Use approvals or explicit confirmation for externally visible actions.
* Disconnect tools that should no longer be available.