Cliptude CLI + MCP

Generate videos from a saved template. Same account and credits as cliptude.com.

Also from Cliptude: Awesome Faceless YouTube Automation, a curated list of tools, footage sources and YouTube policies for faceless channels.

Set up

1. Install the CLI: run npm i -g @cliptude/cli.
2. Authenticate: run cliptude auth login and complete the sign-in in the browser it opens.
3. Install the companion skills: run npx skills add cliptude/cliptude.

From this repo without npm:

pip install -e .
cliptude auth login

Cliptude desktop app

The desktop app ships this CLI and MCP server, so nothing needs installing: in the app, open Settings → Connect an AI assistant and click Add for Claude Desktop or Cursor, or copy the command for Claude Code or Codex. Videos are then made in the app with your own provider keys. ChatGPT only connects to servers on the internet, so it uses cliptude.com.

Where the CLI and MCP send jobs:

CLIPTUDE_TARGET overrides the saved choice, and CLIPTUDE_BASE_URL overrides both. The app's port changes every launch: the CLI reads it from desktop-runtime.json and its key from local-api.json in the app's data folder. The desktop gets only that local key and cliptude.com only your cliptude.com login. A CLIPTUDE_API_KEY in the environment, without CLIPTUDE_TARGET, means cliptude.com.

Commands

cliptude templates
cliptude template-options --format faceless_documentary
cliptude voices --provider elevenlabs
cliptude template-create --name "TSMC doc" --format faceless_documentary --set voice_id=<id> --set voice_name=George --set generation_model=pro --set faceless_documentary_style=investigative
cliptude template-create --name "TSMC doc" --format faceless_documentary --settings-file settings.json
cliptude generate --template "Finance vector weekly" --title "Why index funds win" --prompt "why most retail traders underperform the S&P over 10 years, keep it for beginners" --duration 3-4
cliptude estimate --template "Finance vector weekly" --script-file script.txt
cliptude avatars --section mine
cliptude generate --template "My avatar weekly" --title "Episode one" --script-file script.txt --avatar 42
cliptude generate --template "Finance vector weekly" --title "Why index funds win" --script-file script.txt
cliptude generate --template "Finance vector weekly" --title "Why index funds win" --voiceover narration.mp3
cliptude generate --template "My talking head" --title "Weekly recap" --talking-head clip.mp4
cliptude generate --template "My avatar weekly" --title "Grandma egg trick" --script-file script.txt --avatar-ai-footage --avatar-intro-only
cliptude script --template "Finance vector weekly" --title "Why index funds win" --prompt "why most retail traders underperform the S&P" --duration 3-4
cliptude voiceover --template "Finance vector weekly" --title "Why index funds win" --script-file script.txt
cliptude download-voiceover <job-id> --out narration.mp3
cliptude editor <job-id>                                   # editor project id, revision, timeline item ids
cliptude edit <project-id> --ops '-' <<< '[{"op": "delete", "item_id": "anim-agent_g10"}, {"op": "retime", "item_id": "anim-agent_g11", "start_s": 109.3}]'
cliptude agent <project-id> --prompt "Replace this shot with real footage of a TSMC fab" --items clip-1 --tier smart --wait
cliptude render <project-id> --wait                        # new MP4 of the edited timeline
cliptude continue <job-id>
cliptude status <job-id>

No template yet? template-create makes one without the create wizard: pick a voice_id from voices, and see template-options for every other field, its choices and its default (unset fields take the wizard default). Values are validated like the wizard; uploads (custom music, avatars, characters) still need the wizard. MCP tools: get_template_options, list_voices, create_template. For Avatar Video, avatars (MCP list_avatars) lists your avatars; set one with --set avatar_profile_id=<id> on template-create, or override it per video with --avatar <id> (and --avatar-engine iv|v) on generate, plan-job and estimate. The avatar's gender must match the voice. For Vector Explainer, vector-characters lists the built-in cast and your uploads, vector-character-upload --name "Dr. Maya" --file maya.png (or --url) adds your own, and --set vector_character_id=<id> on template-create (or --vector-character <id> on generate / plan-job) uses it. Motion: --set vector_reveal_mode=none|video on the template, or --vector-motion none|video on generate, plan-job and estimate (none = stills with camera moves, video = AI clips). Thumbnails: --thumbnail-reference <YouTube link or image URL> or --thumbnail-prompt "<brief>" on template-create sets the template's look; the same flags on generate and plan-job set one video's. With neither, each video picks a style reference for its own title.

To check the price first, estimate (MCP estimate_credits, REST POST /api/v1/jobs/estimate) takes the same template and prompt/script as generate (an upload as --voiceover-seconds / --talking-head-seconds) and returns the credits the job would need and whether the account has them. It creates and charges nothing.

To revise a finished video, editor opens its timeline (item ids, times, text). edit <project-id> --ops ops.json applies exact edits on those ids, free and in one revision: delete, retime, trim, set opacity/volume, set a graphic's props, design or picture (an uploaded keep id or https link), mute a graphic's sound, add a graphic, clip or catalog sound effect, set a transition, restyle captions, set library music, its volume or remove it, set the background. versions and restore undo. agent asks Cliptude Agent for open-ended edits that need sourcing or generation (it charges agent credits; reference item ids with --items), and render exports a new MP4. MCP tools: open_editor_project, get_editor_project, edit_timeline, list_editor_versions, restore_editor_version, run_agent, get_agent_run, cancel_agent_run, render_project, get_render.

Agent-planned videos

On Faceless Documentary, Faceless Top 10, Avatar Video and Vector Explainer templates an agent can plan every visual and leave Cliptude only the voicing, sourcing and rendering:

1. cliptude plan-job (stops at plan_ready), cliptude narration <job>. Like every job, it answers first with a review of its settings; music, sound effects, transitions, animations and web sources are on by default, and any the template turns off (no_music, no_sound_effects, no_transitions, no_motion_graphics, ...) need the user's approval in --acknowledge. Your own footage: cliptude upload-media <job> trailer.mp4 --split-shots cuts a whole source into checked 3-12 s clips on its cuts and prints their keep ids; each upload's notes say what the build review would report (a watermark, an end screen, captions).
2. cliptude catalog --compact, then --types a,b for full entries (--premium lists only the premium boards).
3. cliptude plan-check <job> plan.json until valid, cliptude plan-lock <job> plan.json, cliptude continue <job>. The job stops at ready_for_edit: built and charged, not rendered.
4. cliptude plan-review <job>: every picture or clip (asked vs used, link to the file) and every graphic (the props it will be drawn with, or why it is missing). cliptude plan-report <job> adds each fallback.
5. To revise, plan-check / plan-lock a new plan on the same job and continue again: same narration, unchanged media reused, nothing more charged.
6. Render once: cliptude editor <job>, cliptude render <project> --wait.

Aim for two builds: one to review, one to ship. The skill's "One pass" section has the order of work (footage budget, sources split by shot, people's photos first, picture assignment by script, person-card caps).

MCP tools: create_plan_job, get_narration, get_kind_catalog, check_plan, lock_plan, get_plan, continue_job, get_plan_review, get_plan_report, then open_editor_project and render_project. The plan sets the look (theme, style, colours, music), every graphic with its own data, colours and sound, the footage under every beat, the transition on each cut, two-image reveals, and free sound effects; get_kind_catalog lists the graphics, transitions, reveal_styles, clip_effects and sounds (with previews). The plan schema and rules are in skills/cliptude-plan/SKILL.md. The job costs the same as the website price.

Product videos from a store

Turn a store product into a 6 to 60 minute educational YouTube video: about 80% teaching, then the product, with its real photos composited into the closing boards (never regenerated). Any Shopify product URL works with no setup; other stores work through the page's structured data; a store connected at cliptude.com/integrations/stores/ (or that installed the Cliptude Shopify app) can also be searched.

cliptude product fetch https://store.com/products/handle        # product_id + 3 angles
cliptude product script <product-id> --angle 1 --duration 10-12 --discount-code YOUTUBE15
cliptude product status <job-id>                                 # until script_ready; review the script
cliptude product render <job-id> --aspect-ratio 16:9             # returns at once
cliptude product status <job-id>                                 # until ready_for_review; open editor_url

MCP tools: fetch_product_context, draft_product_script, render_product_video, get_product_video_status. None of them waits for a render; poll the status tool.

--title is required. Provide one of --prompt, --script / --script-file, --voiceover, or --talking-head. Never copy the title into --prompt. If you only have a topic, follow cliptude-script (web search → spoken narration → --script-file), or ask Cliptude to write it with cliptude script. After a completed script or voiceover job, cliptude continue <job-id> turns that same job into a full video (reuses the script; voiceover jobs reuse the TTS file). --voiceover is a local audio file (same as create.html); it is transcribed and skips TTS — do not combine it with prompt/script. --talking-head is a presenter video on a Talking Head template (MP4, MOV, WEBM, AVI); it is transcribed and skips TTS. Optional --duration is a wizard bucket (3-4, 6-8, 10-12, …). Omit it to use the template duration, or the word-count estimate when you pass --script. Uploaded audio or talking-head video sets duration from the file. Avatar Video templates can override Avatar Intro Only and AI Image + AI Video with --avatar-intro-only / --no-avatar-intro-only and --avatar-ai-footage / --no-avatar-ai-footage. Omit those flags to keep the template. generate / script / voiceover / continue return immediately (queued). Poll status.

MCP

Remote Streamable HTTP: https://mcp.cliptude.com/mcp (OAuth or Bearer).

cliptude-mcp

Unauthenticated POST /mcp returns 401 with WWW-Authenticate pointing at RFC 9728 metadata. Stdio never writes to stdout except JSON-RPC.