AutoClip MCP Server: Setup for Claude Code, Cursor and Other Apps

• Turns a long video or link into short clips for chosen platforms

  Starts a background job that makes videos, covers and captions for platforms such as douyin, tiktok, reels or shorts. The source can be a local file or a YouTube or Bilibili link.

  Try: “Make TikTok and Shorts clips from ~/Videos/interview.mp4.”

  Runs code · start_quick_output

• Checks progress and collects the finished clips

  Shows progress and the finished videos, covers and captions for each platform. Ask for the publish kit to get a ZIP.

  Try: “How far along is project abc123? Give me the publish kit when it is done.”

  Files · get_quick_output_status

• Cuts a local video into clips and waits until it is done

  Runs the older full clipping job on a local video and reports progress. It can take minutes to tens of minutes. You can pick the model provider.

  Try: “Clip ~/Videos/talk.mp4 and keep only the best segments.”

  Runs code · clip_video

• Starts the older clip job in the background

  Starts clipping and returns a project ID straight away, for clients that time out on long calls.

  Try: “Start a clip job for ~/Videos/talk.mp4.”

  Runs code · start_clip_job

• Checks an older clip job

  Reports progress, stage and estimated time for a job started with the background clip tool.

  Try: “What is the status of project abc123?”

  Read · get_job_status

• Shows the clips of a finished project

  Returns the clips with titles, times, scores and file paths, plus the output folder.

  Try: “List the clips from project abc123.”

  Files · get_project

• Lists recent projects

  Lists recent AutoClip projects. It shares the data folder with the desktop app.

  Try: “Show my last 10 AutoClip projects.”

  Files · list_projects

• Shows which AI models it can use

  Lists the cloud model providers and local presets such as Ollama and LM Studio, and which one is active.

  Try: “Which model providers can AutoClip use?”

  Read · list_providers

• Checks that ffmpeg and the model are ready

  Checks ffmpeg, the Whisper speech tool and the model connection before you start a job.

  Try: “Check that AutoClip is set up correctly.”

  Runs code · check_environment

• Renders one clip ready to post

  Renders a clip as a vertical 9:16 video with burned-in subtitles and a title card, using a platform preset. The same settings reuse a cached file.

  Try: “Export clip 3 of project abc123 for Shorts.”

  Runs code · export_clip

• Posts a clip to social platforms

  Renders a clip and posts it through the Upload-Post service to platforms such as TikTok, Instagram, YouTube, Facebook and X. It can schedule a post. It needs an Upload-Post API key and profile.

  Try: “Post clip 3 to TikTok as private.”

  Write · publish_clip

• Checks whether a post went live

  Shows whether each platform post is pending, done or failed, with links.

  Try: “Did my last post go through?”

  Network · get_publish_status

• Lists your Upload-Post accounts

  Lists the Upload-Post profiles and connected platforms, and whether publishing is set up.

  Try: “Which accounts can I post to?”

  Network · list_publish_profiles

• Stops a running render

  Stops the ffmpeg work for one project, clears its temporary files and marks the job cancelled. Videos that already finished are kept.

  Try: “Cancel project abc123, it is taking too long.”

  Runs code · cancel_job

• Lists the editing styles you can ask for

  Returns the editing styles the template option accepts: editorial, street and classic.

  Try: “Which clip styles can I choose?”

  Read · list_styles

• Used by the assistant behind the scenes

  Shows the installed version

  Returns the AutoClip CLI and MCP version number.

  Try: “Which AutoClip version is installed?”

  Read · get_version

Watch out for

• It can post videos to your social accounts

  The publish tool sends clips to platforms through Upload-Post. Ask the assistant to use private or scheduled posts first, and only set the Upload-Post key when you want posting. github.com

• It writes video files to disk and runs FFmpeg

  Jobs create projects and rendered files in the AutoClip data folder. Check the output folder size, since long videos take space. github.com

• Cloud models see your transcript

  With a cloud provider, text from your video goes to that provider, and cloud models are billed by use. Use a local model such as Ollama to avoid this. github.com

• Sends anonymous usage events

  Tool name, client name, version, run time and success are sent to PostHog. Paths, file names and keys are not. It is off when analytics is false in privacy.json. github.com

Security advisories