CueFrame

What CueFrame MCP server does

CueFrame MCP server gives compatible agents access to a hosted video-editing workflow. It works with editable projects that combine source footage, captions, motion graphics, audio, timing, and other layers. The resulting project can be previewed, revised, and rendered as an MP4 rather than producing only a single flattened output.

The service supports tasks such as turning interviews into short vertical clips, changing an existing title while preserving the rest of an edit, creating social or brand videos, and adapting one edit to additional formats. The repository also includes agent skills that provide workflow guidance for storyboarding, social reels, product videos, motion graphics, music beds, and quality checks.

How it works

An agent sends a request and supplies the required media or project context. CueFrame’s hosted MCP endpoint then supports account checks, media and project preparation, previews, cost estimates, and render submission. Long-running work returns an identifier; the agent waits for completion and reports the result, including an MP4 link and project ID when available.

Projects remain editable. For example, an agent can preview a template-based edit, replace one title, and preserve the footage, captions, audio, timing, and placement. Cloud operations such as uploading footage and rendering consume credits, so the workflow can check available credits and estimated costs before paid work begins.

Setup and configuration

The CueFrame MCP server is available at https://api.cueframe.ai/v1/mcp over Streamable HTTP. Interactive clients can authenticate through browser sign-in and OAuth. Headless clients use a bearer token in the Authorization header. The material does not specify a required environment variable for either setup.

For local agent setup, Node.js 24 or newer is required by the npx -y cueframe@0.8 install installer. It detects supported agents, adds the hosted endpoint to their configuration, and installs skills in the user’s agent skill directories. Use --dry-run to inspect planned changes. Manual MCP configuration is also supported, and the repository provides client-specific examples.

Claude Desktop and ChatGPT use their hosted connector setup rather than the terminal installer. Claude Code and Codex also have plugin installation paths, while Gemini CLI can install the repository as an extension. After connecting, restart the agent session and authenticate when prompted.

Tools and capabilities

CueFrame MCP server supports the following documented workflow capabilities:

• Check the account connection, credit balance, and available credits with get_account.
• Prepare source media and editable video projects.
• Start from templates and modify project layers.
• Preview an edit or a proposed change before rendering.
• Estimate cloud costs before paid operations.
• Submit render jobs and wait for MP4 results.
• Preserve selected parts of an existing edit while changing a specific layer.

Agent skills are separate from the MCP connection. They explain how to carry out supported video workflows, but installing skills alone does not connect an agent to the hosted endpoint.

Limitations and notes

The service is hosted, and cloud operations use credits. Uploading footage and each render can incur charges; the example workflow renders two versions and bills the renders separately. Check the account and estimated cost before starting paid work.

The terminal installer does not configure ChatGPT. The repository’s glama/ shim executes no tools and cannot make videos. Interactive clients can authenticate in a browser, but headless clients need a bearer token. If tools do not appear after setup, restart the client and check its MCP connection settings.