screenverb-mcp

# ScreenVerb MCP Server

Hosted remote [Model Context Protocol](https://modelcontextprotocol.io/) server for [ScreenVerb](https://screenverb.com), a screenshot app for Mac, iPhone and iPad.

The server lets AI agents find, read and organize the screenshots you save to your ScreenVerb cloud library, including the text inside them. Pull an error message, a design mock or an API response out of a screenshot without leaving your editor or chat.

This repository holds the documentation and registry metadata. ScreenVerb hosts and runs the server; there is nothing to install or run locally.

- Server URL: `https://mcp.screenverb.com/mcp`
- Transport: Streamable HTTP
- Auth: OAuth 2.1 with PKCE (no API key)
- Registry name: `com.screenverb/screenverb`

## Before you connect

You need a ScreenVerb account with cloud library access. Sign up at https://screenverb.com.

## Configure

### VS Code with GitHub Copilot

Install from the MCP server gallery in the Extensions view (search for **ScreenVerb**), or add it to `.vscode/mcp.json` in your workspace or to your user `mcp.json`:

```json
{
  "servers": {
    "screenverb": {
      "type": "http",
      "url": "https://mcp.screenverb.com/mcp"
    }
  }
}
```

Start the server from the `mcp.json` editor or the **MCP: List Servers** command. VS Code opens a browser window for ScreenVerb sign-in the first time it connects. Then open Copilot Chat in Agent mode and the ScreenVerb tools are available.

### Other MCP clients

Any client that supports remote Streamable HTTP servers with OAuth can connect to the same URL. A typical config:

```json
{
  "mcpServers": {
    "screenverb": {
      "type": "http",
      "url": "https://mcp.screenverb.com/mcp"
    }
  }
}
```

The client discovers the OAuth endpoints from `https://mcp.screenverb.com/.well-known/oauth-protected-resource/mcp` and registers itself (Dynamic Client Registration and Client ID Metadata Documents are both supported). There is no client ID or secret to configure.

## Sign-in and permissions

When you connect, ScreenVerb asks you to sign in and choose what the agent may do:

- **Read**: list, search and open screenshots (`screenshots:read`)
- **Update**: rename and remove screenshots (`screenshots:write`)

The agent only sees the tools for the access you grant. Tool calls run as the ScreenVerb user who authorized the connection. Revoke access at any time from the MCP page in your ScreenVerb dashboard.

## Tools

| Tool | What it does | Access |
| --- | --- | --- |
| `list_screenshots` | List the newest screenshots with names, dates, captions and tags | Read |
| `search_screenshots` | Search filenames, recognized text, captions, tags and labels | Read |
| `get_screenshot` | Open one screenshot to see the image, its OCR text and metadata | Read |
| `rename_screenshot` | Rename a screenshot | Update |
| `delete_screenshot` | Remove a screenshot from the library | Update, destructive |

`delete_screenshot` is annotated as destructive, and the server tells agents to ask you before calling it. The hosted server is the source of truth for tool names and schemas.

## Usage

Ask the agent in plain language, for example:

- Find the screenshot of the stack trace I took this morning and fix the error it shows.
- Search my screenshots for the expired token error and tell me what it says.
- Open my latest screenshot of the settings mock and build that layout.
- List my 10 most recent screenshots and summarize each one.
- Rename my latest screenshot to Q3 roadmap draft.

## Links

- Docs: https://screenverb.com/docs/mcp
- Website: https://screenverb.com
- Privacy: https://screenverb.com/privacy
- Terms: https://screenverb.com/terms
- Support: support@screenverb.com

## License

MIT