Yes, Claude or ChatGPT can make an animated explainer video for you through ClipOnIt's MCP server: you give a topic, a text or your voice, the assistant calls explainer_produce, waits with explainer_job_wait and hands you the link to the vertical (9:16) MP4, with voice-over and captions. The MCP server and the API work from the free plan (3 watermarked videos a month); Starter ($29/month) gives 15 videos a month without a watermark. What stays with you: picking the topic, checking the facts, and confirming each TikTok post.
What is an MCP server, in one sentence?
The Model Context Protocol (MCP) is an open standard that lets an AI assistant (Claude, ChatGPT, Cursor…) call a third-party app's functions as if they were its own tools. The app exposes a list of named tools with their parameters; the assistant discovers them and chains them to answer a plain-language request. For the user, a point-and-click interface becomes a conversation: "make me a 40-second video on the history of the barcode, cartoon style, and give me the link".
Why go through an assistant rather than the interface?
ClipOnIt's interface already makes a video in one click. The assistant helps when the video is one step of a bigger job:
- Series: ten topics on the same theme, back to back, in the same style and voice.
- Research first: the assistant reads an article, pulls the key facts, writes the narration, then starts the video in the same conversation.
- Product or agency: one video per release note, product page or client, triggered from Claude Code or a script.
For an occasional video, the web interface is enough. Make a few videos by hand first to find your style; automate afterwards.
The full flow, tool by tool
- Connect. With Claude Code: claude mcp add --transport http cliponit https://cliponit.app/mcp --header "Authorization: Bearer ck_…" (key created in the app, Settings → API). With Claude Desktop or claude.ai: a custom connector on https://cliponit.app/mcp, signed in through OAuth. With ChatGPT: a remote MCP through the Responses API, or a GPT Action from the OpenAPI file.
- Pick a style. explainer_styles lists the 24 visual styles (2D and 3D), plus your own if you imported one. Without a choice, the assistant uses pixel-news.
- Estimate. explainer_produce with dry_run: true checks the request and the quota without creating or billing anything: renders left, estimated length, AI budget reserved.
- Produce. explainer_produce with one source: topic (AI writes the narration), script (a ready script, no AI cost) or audio_url / audio_id (your own voice, transcribed then illustrated). The server writes, checks and renders the video in the background and returns a job id right away.
- Wait. explainer_job_wait blocks until the end (count 3 to 8 minutes) and returns url, the link to the 1080×1920 MP4, and editor_url, which opens the video and its version history in the editor.
- Tweak. If a scene is off, open the editor and ask the AI Director for the change, or let the assistant use the other explainer_* tools (scene variants, new render).
- Post. The video posts to TikTok from ClipOnIt once your TikTok account is connected. TikTok requires the account owner to confirm every post: the assistant prepares, you approve.
The server exposes 45 explainer_* tools in total (element library, scripts, voices, variants, templates, brand kit), plus get_quota to see what is left this month.
The demo below shows this flow for real, sped up: Claude is asked for a video about the fall of Rome, calls explainer_produce, waits with explainer_job_wait, then gives the MP4 link.
A prompt that works
"With ClipOnIt, make a 40-second explainer video on why the sky is blue, for 12-15 year olds. Style croquis-cours. Estimate first, then start the production, wait for it and give me the video link and the editor link. Also suggest a TikTok caption of 100 characters max."
Three details change the result: give a length (otherwise the narration grows), name the audience (the vocabulary follows), and ask for the caption in the same message (the assistant has the narration in front of it).
If the text is already written, send it as script rather than topic: it is read as is, just like in the interface (see the script to video page).
REST API or MCP: which one?
| REST API (script) | MCP server (assistant) | |
|---|---|---|
| Who uses it | A developer | Anyone who can write an instruction |
| When | Same flow, repeated, built into another system (CMS, back office) | A different topic each time, research and writing in the same conversation |
| Setup | A few hours of code (POST /api/agent/produce, then follow the job) | A few minutes of configuration |
| Risk | Rigid: changing the instruction takes code | The assistant can misread; ask for the estimate before producing |
Both run on the same API, described in OpenAPI 3.1 at cliponit.app/openapi.json. curl, JavaScript and Python examples on the integrations page.
What does not automate
- Picking the topic. AI writes what you ask for; knowing what your audience cares about is still your job.
- The facts. Read the narration before posting, especially on numbers or news.
- Confirming the post. An API key cannot post to TikTok on its own: the account owner approves each post.
Security and data
An API key gives access to your account; treat it like a password (environment variable, never in a shared prompt). ClipOnIt hosts files on Hetzner servers in Finland.
FAQ
Can Claude make a video on its own?
Not natively. It needs a video tool connected through MCP or an API. ClipOnIt writes the scenes, animates them, adds the voice and renders the video; Claude orchestrates and hands you the link.
Do I need to code?
Not for MCP: a URL, a sign-in and plain-language instructions. Yes for the REST API.
Which plan do I need?
The MCP server and the API work from the free plan (3 watermarked videos a month, 8 styles), then Starter ($29/month, 15 watermark-free videos a month), then Pro ($49, 30 videos) and Business ($149, 100 videos). The free plan gives you 3 videos a month, with a watermark, from the web interface. Details on the pricing page.
Can I use my own voice?
Yes: give an audio file to explainer_produce (audio_url or audio_id). The voice is transcribed word by word and each element appears when you name it. Good lengths are in short video length and format.