Zapcap MCP Server

What bogdan01m/zapcap-mcp-server MCP server does

The bogdan01m/zapcap-mcp-server MCP server exposes selected ZapCap video-processing operations as MCP tools. It can send a local video file or a video URL to ZapCap, retrieve the processing templates available from the service, create a task, and check that task’s progress.

Task creation supports caption and presentation settings rather than only a basic processing request. Agents can select a template, choose a language, enable automatic approval, and optionally turn on B-roll with a configurable percentage. Subtitle controls cover emoji, emoji animation, keyword emphasis, animation, punctuation, and the number of displayed words. Visual options include position, casing, font size, weight, colors, shadow, stroke, and several highlight colors.

This is an unofficial ZapCap MCP implementation. It is intended for workflows where an MCP-compatible client should coordinate video preparation without manually constructing each ZapCap API request.

How it works

The client starts the server as a local stdio process. The server reads the ZapCap credential from its environment and uses it for the available API operations. A typical workflow is:

1. Upload a video from a file path or URL.
2. Use the returned video identifier when creating a task.
3. Select a template and provide caption, B-roll, language, and styling parameters.
4. Use the video and task identifiers to monitor progress.

The repository documents automatic parameter validation through Pydantic, including type checking for tool inputs. Natural-language requests can be translated by the MCP client into the structured arguments required by these tools.

Setup and configuration

The bogdan01m/zapcap-mcp-server MCP server requires uv and a ZapCap API key. The documented MCP configuration runs the PyPI package with uvx and sets ZAPCAP_API_KEY in the server environment:

{
  "mcpServers": {
    "zapcap": {
      "command": "uvx",
      "args": ["zapcap-mcp-server"],
      "env": {
        "ZAPCAP_API_KEY": "your_api_key_here"
      }
    }
  }
}

The README specifically gives Claude Desktop and Cursor as example MCP clients. An alternative installation uses uv tool install zapcap-mcp-server. Docker is also documented through the published bogdan01m/zapcap-mcp-server:latest image, with the API key passed into the container.

The key must be obtained from the ZapCap platform. Keep it in the environment rather than placing it in individual tool arguments.

Tools and capabilities

The server documents these tools:

• zapcap_mcp_upload_video: uploads a video from a local file_path.
• zapcap_mcp_upload_video_by_url: sends a video identified by url to ZapCap.
• zapcap_mcp_get_templates: lists available ZapCap processing templates.
• zapcap_mcp_create_task: creates a task using a video_id, template_id, and caption, B-roll, language, and styling options.
• zapcap_mcp_monitor_task: checks progress using a video_id and task_id.

B-roll is optional and can be enabled with a percentage from 0 to 100. The documented defaults include English, automatic approval, animated subtitles, keyword emphasis, and one displayed word, although callers can override the supported settings.

Limitations and notes

The server depends on ZapCap and therefore requires a ZapCap API key. The implementation is unofficial, so behavior depends on the ZapCap API and the package’s supported tool parameters. The README describes testing integrations, named configurations, and template-default overrides as future plans rather than current capabilities. No claim is made here that those planned features are available now.