API & MCP

Build a transcript workflow for an API or MCP client

Understand Pyntic's transcript endpoint, source boundaries and the additional work required to expose a real MCP tool.

Pyntic teamReviewed 2 min

An HTTP transcript endpoint and an MCP server solve related but different integration tasks. An endpoint returns data; an MCP server exposes tools through a defined protocol to compatible clients.

Start with the HTTP contract

Pyntic exposes GET /api/transcribe?url=VIDEO_URL with an optional lang parameter. Successful responses include transcript text and segments with start and duration in seconds. Check HTTP status and error fields rather than assuming every public video has captions.

Preserve a stable data model

Store the source URL, language when known, text and timing together. Do not confuse seconds with milliseconds. When importing data from another service, normalize units before writing SRT, VTT or timestamp links.

Expose a bounded tool

For an MCP integration, wrap the endpoint in a tool with a clear URL schema, response contract and error handling. Authenticate callers, limit request size and rate, and treat retrieved captions as untrusted data rather than instructions.

Configure the client explicitly

Claude Desktop, Cursor and other clients need the transport and server configuration supported by their installed version. Pyntic exposes a stateless Streamable HTTP MCP endpoint at /api/mcp. The operator must configure MCP_ACCESS_TOKEN; clients use its Bearer authorization header. Check client transport support before connecting.

Sources and documentation

Put it into practice

Work with your own content.

Open the tool →

Keep learning