> ## Documentation Index
> Fetch the complete documentation index at: https://docs.mka1.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Connectors, skills, and plugins

> Connect external tools and manage reusable instructions for coding tasks.

export const ScreenshotCrop = ({src, alt, width, height, x = 0, y = 0, cropWidth = width, cropHeight = height, maxWidth = "100%"}) => <div className="not-prose" style={{
  position: "relative",
  overflow: "hidden",
  width: "100%",
  maxWidth,
  margin: "0 auto",
  aspectRatio: `${cropWidth} / ${cropHeight}`
}}>
    <img src={src} alt={alt} width={width} height={height} style={{
  position: "absolute",
  display: "block",
  margin: 0,
  maxWidth: "none",
  width: `${width / cropWidth * 100}%`,
  height: "auto",
  left: `${-x / cropWidth * 100}%`,
  top: `${-y / cropHeight * 100}%`
}} />
  </div>;

Add tools and reusable instructions when a task needs more than the files in your project. Connectors expose operations from external services; skills provide instructions, and compatible plugins package supported components.

## Choose the right extension

| Component | What it provides | Example |
| - | - | - |
| Connector | Tools that call a configured service or local MCP server. | Read a bug report from an issue tracker. |
| Skill | Reusable instructions for a type of work. | Follow a document-generation or review workflow. |
| Plugin package | A bundle of supported components. | Install a package containing skills and connectors. |
| Marketplace | A source from which packages can be discovered. | Add a team's package repository. |

A skill does not grant access to a service by itself. If its workflow needs an external tool, configure the corresponding connector and account access too.

## Connect external tools

A connector gives the coding agent tools from an external service through the Model Context Protocol (MCP). Connect your issue tracker to use a bug report's reproduction steps, or a design service to bring specifications into an implementation task. The available operations depend on the server and the access granted to your account.

1. Open **Settings → Extensions** and choose a connector from the catalog. Use **New server** if you need to configure a custom MCP server.
2. Complete the connector's sign-in or credential setup and enable it.
3. Choose **Test** and inspect the discovered tools. A successful test confirms the app can connect and list the server's tools.

After the test succeeds, name the service and the information to use in your coding request. For example, ask the agent to read a checkout bug report and use its reproduction steps to add a regression test. The agent's calls appear in the timeline and follow the applicable permission policy.

### Add a custom MCP server

Open **New server** when your service is not in the catalog. Use the transport and configuration supplied by that server's documentation:

| Field | What to provide |
| - | - |
| **Server name** | A unique name you can recognize in the tool list. |
| **Transport** | HTTP or SSE for a URL, or stdio to launch a local process. |
| **URL** | The server address for HTTP or SSE. |
| **Command** and **Arguments** | The local executable and one argument per line for stdio. |
| **Working directory** | The folder the local server should run in, if required. |
| **Headers** or **Environment variables** | The name-value pairs required by the server. |
| **Timeout (ms)** | A positive timeout override, if the default is unsuitable. |

Save the configuration, complete authorization if requested, and run **Test**. For stdio, the executable and its dependencies must exist on the machine running the connector. A successful test confirms connection and tool discovery; it does not guarantee that your account can access every resource through those tools.

### Use the connection in a task

Name the service and the specific input rather than asking the agent to search everything:

> Read the checkout quantity bug in the connected issue tracker. Summarize its reproduction steps, identify the relevant code, and propose a regression test. Do not modify the issue or project files yet.

Inspect the connector's tool results in the timeline. If the requested resource is unavailable, verify account access and the resource identifier. If the tools themselves are missing, check that the connector is enabled and repeat **Test**. Configuration changes apply to the next run; an active run keeps its existing configuration.

<Frame caption="The connector catalog shows each service and its connection status.">
  <ScreenshotCrop src="/images/mka1-code/guides/extensions.jpg" alt="GitHub, Linear, and Notion connectors with Connect buttons" width={2720} height={1660} x={860} y={350} cropWidth={1250} cropHeight={350} maxWidth="100%" />
</Frame>

## Skills and plugins

**Extensions** also manages skills and plugin packages. A skill supplies instructions for a kind of work. A compatible plugin package can include skills, connectors, and other supported components.

In the **Skills** tab, enable or disable individual skills. Open a package in **Packages** to inspect its components. **Marketplaces** lists package sources; adding a source can make more packages available to install.

This configuration is shared by local desktop, VS Code, and CLI sessions that use the same local store.

### Install and inspect a package

1. Open **Settings → Extensions → Packages** and inspect an available package's source and components.
2. Install the package, then check its components in **Skills** and **Connectors**.
3. Enable the components you need and complete any connector authorization.
4. Start a new run and request a small task that exercises the intended workflow.

Adding a marketplace only makes its packages discoverable; it does not connect every service they contain. In **Marketplaces**, enter an `owner/repo`, HTTPS URL, or local path under **Marketplace source**, then choose **Add**. Use the **Skills** switches to control which reusable instructions are available to the agent.

Local desktop, extension, and CLI sessions can share this configuration through the same store. A remote workspace has its own environment; see [tools and credentials for remote workspaces](/docs/mka1-code/workspaces#remote-environments) before expecting a local connector or skill to work there.

Continue with [Quickstart](/docs/mka1-code/first-task).