Slicer MCP server
Connect your AI assistant to Slicer over MCP — it edits a vertical video itself with editor commands, for free. Only Slicer's AI features are paid.
On this page
MCP (Model Context Protocol) is an open protocol AI assistants use to connect to external services. The Slicer MCP server gives your assistant — any client that supports MCP — the whole Slicer editor: upload materials, build a video from them, polish captions and elements, export and publish. You just tell the assistant what you need:
Make a reel from
morning.mp4andlatte.jpg: start with the photo, add the hook "Morning at the coffee shop" and soft transitions.
The assistant uploads the files, builds the video itself and sends you a link to the project. All of it is free.
Your assistant edits the video itself
Your assistant does the editing — just like you do in the app with Edit it myself. By default, create_project
creates exactly that kind of project: the materials go onto the timeline in order, and then the assistant builds the video
with editor commands — the same operations as the buttons in the app:
- Uploads the materials — local files via a one-time link, links via import.
- Creates the project — free, with no price and no confirmation; within seconds the project becomes a draft.
- Reads the video as text — fragments, their timing, elements.
- Builds the video — fragment order and trimming, hook, titles, plaques, zooms, emoji, transitions, colour, music and caption style.
- Exports and publishes — rendering the file and publishing to connected social accounts are free too.
If the materials contain speech, the assistant offers captions from speech: it's a paid AI feature, so it names the price first. Without it such a project has no captions — everything else works anyway.
Your own director
The assistant can also edit the video itself, like a director: get_director_brief gives it the material by words and
the plan schema, get_frames gives it frames of the sources, and submit_director_plan accepts the plan together with
what the assistant saw in the frames. The rules of the Slicer AI director are applied to the plan on our server: the
assistant gets not the rules but fixes and issues with its plan, fixes it and submits it again. All of it is free, and
the direct_reel prompt sets the workflow — Your own director.
What's free and what's paid
| Free | Paid — Slicer's AI features |
|---|---|
creating a project (create_project, by default) | AI montage by the Slicer director — create_project with mode: "ai", rebuild remount_project, retry retry_project |
reading the video (get_transcript); your own director — brief, frames and accepting the plan | captions from speech — transcribe_speech, at cost per minute of sound |
editing — the whole operation catalog (edit_project), style (set_style), auto colour (auto_color) | the Slicer AI assistant — ask_assistant, at cost |
export and publishing (export_video) | caption translation — translate_captions, at cost |
| uploads, project list and status, choosing a version, renaming, deleting, docs search |
An AI montage costs 24 tokens per video up to 90 seconds; the first video is free. The assistant only offers paid features and never starts them on its own: the AI director — only if you ask for it.
Money only with your consent
A paid tool first returns the price and starts nothing. To start, the assistant has to show you the price and repeat the
call with confirmCost equal to it. Captions from speech, the AI assistant and translation are priced at AI cost: the
quote has an expected and a maximum amount, and the actual price is charged, never more than the confirmed one. On top of
that, the connection has a spend limit: 200 tokens a month by default, and you set it on the consent screen. Details —
Security and costs.
Requirements
| What | Value |
|---|---|
| Server URL | https://api.slicer.media/mcp |
| Transport | Streamable HTTP, sessionless: every request is self-contained |
| Sign-in | OAuth 2.1 — sign in to Slicer in a browser and approve on the consent screen; or an Authorization: Bearer slc_… header with an API token |
| Protocol versions | 2026-07-28, as well as 2025-11-25, 2025-06-18 and 2025-03-26 — every current client works |
| Account | a Slicer account; editing by your assistant is free, only Slicer's AI features spend tokens |
How to connect the server in different clients — step by step on the Connecting page.
What the assistant can do
| Group | Tools |
|---|---|
| Account and catalog | balance and prices, styles and montage settings, price estimates |
| Uploads | a link for uploading a file, uploading from a public URL, status |
| Projects | create a project — free by default, without the AI director; list and check projects, choose a version, rebuild, retry, stop, rename, delete |
| Editing | the video text with word ids, edit operations — from cuts and titles to music, markers and the cover, style, auto colour, captions from speech, AI assistant, caption translation |
| Export | rendering with file settings and publishing to connected social accounts |
| Docs | searching and reading these docs — the assistant finds answers about Slicer on its own |
The full list with parameters — Tools. Ready-made conversations — Examples; the
make_reel, edit_reel, direct_reel, edit_myself and publish_reel prompts start them in one command — Prompts.
The finished video
The assistant sends you a link to the project in the web app (appUrl) — there you can watch the video, polish it in the
editor and download it. For programs, the video is available at /v1/projects/{id}/result with a token. There are no
public download links without signing in: only you can see your video.