Guide to the Qencode MCP server

Before you start

To use the Qencode MCP server you need:

• An MCP-capable client (Claude, Claude Code, Cursor, ChatGPT, Grok, Lovable or Gemini)
• A Qencode account

API keys are generated and managed per Project in your Qencode account.

See Connect the MCP server to your client for the current list of supported clients.

Transcoding minutes you start through the MCP server bill against your Qencode plan the same way as jobs you start through the API or dashboard. See Plans and pricing.

The Qencode MCP server lets AI tools interact with Qencode using the Model Context Protocol, a standardized interface for AI agents. It helps developers and content teams run video transcoding workflows directly from their AI tools. Your client draws on the full transcoding API and Qencode's encoding knowledge base, so it can turn a plain-language request into a correct, running job.

This page is a high-level overview of the Qencode MCP server. For step-by-step instructions and example prompts, see the supporting guides below.

In this documentation:

• Set up the MCP server per client covers connecting Claude, Claude Code, Cursor, ChatGPT, Grok, Lovable or Gemini to the server.
• Tools and prompts reference covers what you can ask for, plus the parameters and tested prompts for every tool on the server.

With the server connected, you can:

• Start a transcoding job from a prompt: Ask for an MP4, an audio extract, or a full streaming ladder in plain language.
• Build adaptive bitrate streaming (HLS and DASH): Generate a master playlist with per-rendition variant playlists and segments for browser, iOS, and CDN delivery.
• Generate subtitles, transcripts, and translations: Attach subtitle files, preserve closed captions, or auto-generate transcripts from speech in. SRT, VTT, and timestamped JSON.
• Create thumbnails and sprite sheets: Capture thumbnails at a specific time, frame, or interval, let AI choose the best frame with Smart Thumbnails, or assemble thumbnails into sprite sheets for scrubbing previews.
• Stitch and clip source videos: Concatenate, or trim inputs to build compilations, intros, and outros.
• Search and read Qencode's knowledge base: Query Qencode's reference documentation from inside your client and read any document in full.
• Track jobs to completion: You get completion percentage, output URLs, and any warnings or errors without leaving the conversation.
• Manage your Qencode Media Storage: List buckets, browse contents, create buckets, stage inputs, and hand off files with download links.

Set up the MCP server

Connect to the Qencode MCP server through its hosted endpoint:

https://mcp.qencode.com/mcp

No additional software or local installation is required beyond an MCP-compatible client. For step-by-step instructions per client, see the setup guide.

Authentication uses your Qencode API key. You can find and manage the API key for each Project in your Qencode account. Keep your API key private, as it authorizes transcoding jobs billed to your Qencode plan.

Connect the MCP server to your client

Follow the setup guide for your AI client to connect to the Qencode MCP server. Some clients also support a directory listing or one-click connector, reducing or eliminating manual configuration.

If your MCP-compatible client is not listed, follow its documentation for adding a remote MCP server and use the Qencode MCP server endpoint:

https://mcp.qencode.com/mcp

Important behavior

Temporary output storage

When a transcoding job does not include an output destination, its results are saved to temporary storage. Temporary outputs are removed approximately 24 hours after the job completes. Specify a permanent storage destination when you need to retain the files.

Job status and completion percentage

Use the job's status field as the primary indication of whether it has completed. While a job is still queued or processing, its output arrays may arrive empty even though the job is progressing normally, so rely on status rather than the presence of results to judge completion.

Waiting for a job to finish

If wait_for_job returns before the job has reached a terminal state, meaning it timed out rather than completing or failing, check the job once with get_job_status using the same task_token. Use this as a single follow-up check rather than a repeated loop. For jobs that are expected to run long, such as Video Intelligence or wide adaptive bitrate ladders, pass a larger timeout_seconds to wait_for_job instead of polling get_job_status repeatedly.

Temporary download links

Links created by get_download_url remain valid for five to ten minutes. Generate a new link when an existing one expires.

About the knowledge base

The Qencode MCP server uses Qencode's encoding knowledge base to translate natural-language requests into valid transcoding jobs. It provides Qencode-specific information about supported parameters, output formats, and job requirements. When you describe the output you need, the server uses this information to compose a job that follows Qencode's recommended practices.

Need More Help?

For additional information, tutorials, or support, visit the Qencode Documentation page or contact Qencode Support at support@qencode.com.