MCP server
Connect AI assistants to Pipevideo video generation via Model Context Protocol.
Pipevideo provides an MCP (Model Context Protocol) server that exposes video generation tools to AI assistants like Claude, Cursor, and other MCP-compatible clients.
Available tools
| Tool | Description |
|---|---|
list_engines | List available engines and orchestration models |
generate_video | Create a video generation and wait for completion |
get_generation | Check the status of a generation by id |
get_usage | View credit balance and usage statistics |
Connecting
Configure your MCP client to connect to the Pipevideo MCP server endpoint. Authentication uses your Pipevideo API key via OAuth or direct key configuration (depending on your MCP client).
Environment variables
| Variable | Description |
|---|---|
PIPEVIDEO_API_KEY | Your Pipevideo API key (pv_...) |
PIPEVIDEO_API_URL | API base URL (default: https://api.pipevideo.co) |
Example: generate_video
The generate_video tool accepts:
| Parameter | Type | Description |
|---|---|---|
model | string | Orchestration LLM id (e.g. moonshotai/kimi-k2.5) |
input | string | array | Prompt text or Responses-style input items |
instructions | string | Optional system-level guidance |
width | number | Optional canvas width |
height | number | Optional canvas height |
aspect_ratio | string | Optional aspect ratio (e.g. 16:9) |
The tool creates a generation and polls until completion or timeout, returning the video URL or error details.
Use cases
- AI coding assistants — generate video previews from natural language descriptions
- Agent workflows — include video generation as a tool in multi-step agent pipelines
- Prototyping — test video generation interactively without writing API integration code
Related
- TypeScript SDK — programmatic access for your own applications
- Video generation guide — full API flow reference