Search documentation
Search pages, API reference sections, and guide headings.
Summary
The VideoVector MCP docs show how AI clients can browse indexes, run prompts, search media evidence, inspect workflow resources, and validate available tools.
Published package and environment names
Use these package and environment identifiers when configuring MCP clients:
- package:
@videosearch/mcp-server - command:
videosearch-mcp - server name:
videosearch - environment variables:
VIDEOSEARCH_*
Transport choices
- Use
stdiofor local desktop clients such as Claude Desktop and Cursor. - Use
httpwhen the MCP server must run as a network-addressable service.
Pages in this section
- Claude Desktop setup
- Cursor setup
- Custom and HTTP transport setup
- Tools, helper endpoints, and playground
Public helper endpoints
The API also exposes MCP helper routes:
GET /api/v2/mcp/statusGET /api/v2/mcp/configGET /api/v2/mcp/toolsPOST /api/v2/mcp/playground
Use them to inspect capabilities, generate config, and validate tools before deploying the MCP server to users.
Related documentation
This guide connects AI clients to VideoVector tools for media retrieval, prompt execution, workflow inspection, and playground validation.
The public API accepts either API keys or JWT bearer tokens for most workflow endpoints. API key management endpoints require JWT bearer auth, and `/mcp/config` requires a verified JWT session.
Chat sessions provide an agentic retrieval surface on top of search results and prompt-run scope. The API supports session CRUD, turn creation, optional scope narrowing, and streaming turn events over SSE for agent-search experiences.