# Cursor setup

Cursor supports MCP servers, which is the best fit for Slideshot. The editor agent can call Slideshot tools directly from chat, and the CLI is available as a secondary option for scripts you keep alongside the project.

## Copy this into Cursor

Paste this instruction into the Cursor chat panel.

```text
You are helping me connect Slideshot to Cursor.

1. Add the Slideshot MCP server at https://api.slideshot.ai/mcp to my Cursor
   MCP configuration. Walk me through the OAuth approval when the browser opens.
2. Load the Slideshot skill from https://github.com/slideshot/skills into your
   context so you know how to plan a recording and react to paused runs.
3. If shell access is available, you can also call the Slideshot CLI from the
   integrated terminal with `npx slideshot <command>`, using `SLIDESHOT_API_KEY`
   from my shell environment.
4. Create one test recording against a public URL and download `demo.mp4` when
   the run succeeds.

Do not print my API key or any credential.
```

## 1. Add the Slideshot MCP server

Open Cursor's MCP settings and add a server with:

- **Name:** `slideshot`
- **Transport:** HTTP (streamable)
- **URL:** `https://api.slideshot.ai/mcp`

Save the server. On first use, Cursor opens the OAuth approval flow in your browser. After you approve, the Slideshot tools (`create_run`, `get_run`, `list_runs`, `submit_run_input`, `list_run_artifacts`, credentials, brand, feedback) appear in Cursor's tool list.

## 2. Load the Slideshot skill

The fastest install is the shared skills installer:

```bash
npx skills install slideshot
```

Pick Cursor from the list of detected agents. For the manual path, paste the contents of `slideshot/SKILL.md` (from [`github.com/slideshot/skills`](https://github.com/slideshot/skills)) into Cursor's project rules or attach the file to the conversation.

## 3. Use the CLI (optional)

Export your API key once in the shell Cursor runs commands in, then call the CLI through `npx`:

```bash
export SLIDESHOT_API_KEY="sk_live_xxxx"

npx slideshot runs create https://app.example.com \
  --goal "Open the analytics dashboard"
```

## Test the setup

Ask Cursor to record a small workflow against a public URL:

> Create a Slideshot recording of `https://docs.slideshot.ai` that walks through the docs landing page and the api-reference section.

Cursor calls `create_run`, polls `get_run` until status is `succeeded`, lists artifacts, and offers the `demo.mp4` download URL.

## Troubleshooting

- **MCP server not visible in chat.** Make sure the server is enabled in Cursor's settings and the editor was restarted after the change.
- **OAuth approval did not complete.** Check that your browser allowed the redirect back to Cursor. Re-run the approval flow if it timed out.
- **Tool calls fail with `401`.** Re-authorize the MCP connection. The refresh token may have expired.
- **Run paused with `awaiting_input`.** Ask Cursor to read the prompt and submit the requested value via `submit_run_input`.

## Related

- [Install skills](/docs/install-skills): how to install the Slideshot skill across agents.
- [MCP versus CLI](/docs/mcp-vs-cli): when to use the MCP server vs the CLI.
- [Describe the demo flow](/docs/describe-demo-flow): write prompts that produce sharper recordings.