MCP App Store
by KeenEthics
Slipa icon

Slipa MCP Server for ChatGPT

By Salem Ventures

Overview

Slipa turns a rough screen recording into a finished demo video. It transcribes what you said, rewrites it as a clean script, re-voices it with a narrator in any of twenty-nine languages, and renders the result with your logo, brand colours and captions. The narration language is independent of the recording language, so you can record once and publish in another. Ask for a video and Slipa tells you what it will cost before it spends anything.

How to connect

  1. ChatGPT: Open the Slipa listing in the ChatGPT app directory and choose Connect. ChatGPT listing Slipa exposes 22 tools here.

Tools

slipa_create_from_doc

ChatGPT
Turn a written document — release notes, a changelog, a PRD, or any other feature description — directly into a polished, narrated demo video, with NO screen recording involved. Slipa reads the text, writes a script, generates AI narration and on-screen beats, and renders the final video from that alone. Set verbatimNarration (SLP-310, below) to skip the narration-writing step and voice caller-supplied text unchanged instead — the on-screen beats are still generated. Use this when the caller has release notes or a spec and wants a demo video made FROM that text — for a video of an actual screen recording, use slipa_create_upload_url followed by slipa_process (or slipa_create_video) instead. IMAGES ARE SUPPORTED (SLP-301) via the screenshots field below, but NOT as inline binary bytes — a remote MCP client cannot stream those through this protocol. First call slipa_create_image_upload_urls to get presigned S3 URLs, PUT the image bytes to those URLs directly, then pass the returned keys here as screenshots: [{ key, description }]. There is deliberately NO image-by-URL parameter — fetching a caller-supplied URL from this server would be a server-side-request-forgery hole — so a key from slipa_create_image_upload_urls is the only accepted source. SPENDS MONEY: consumes one of the workspace's monthly video-generation slots and real ElevenLabs text-to-speech budget, the same as generating from a recording — only call this with confirmSpend: true when the user has explicitly asked for a video to be created from this document. Poll the result with slipa_get_video using the returned projectId. Resolve any named voice/music/template option with slipa_list_voices / slipa_list_music / slipa_list_templates before calling this — this tool has no brandId parameter (built-in templates only).

slipa_create_image_upload_urls

ChatGPT
Mint one or more presigned S3 URLs to upload SCREENSHOT images for a document-to-video project. Call this BEFORE slipa_create_from_doc when the caller has images to include. The caller must HTTP PUT the raw image bytes to each returned uploadUrl DIRECTLY — not through this tool — with a Content-Type header that exactly matches that entry's contentType. Once every PUT succeeds, pass the returned keys as screenshots: [{ key, description }] on slipa_create_from_doc. Uploading images with this tool ALONE creates nothing — no project, no spend — the video is only created by the subsequent slipa_create_from_doc call, so no confirmSpend is required here.

slipa_create_upload_url

ChatGPT
Start a new video project and get a presigned S3 URL to upload the source screen recording to. A remote MCP client cannot stream a local file through this protocol, so this hands back a URL the caller must HTTP PUT the raw video bytes to DIRECTLY — not through this tool, and not through any other Slipa tool — with a Content-Type header that exactly matches the contentType this call declared (or "video/mp4" if omitted). Once that PUT succeeds, call slipa_process (or slipa_create_video) with the returned projectId to start generation. The URL expires — check expiresIn (seconds) and upload promptly. This call itself does not spend anything; the render/voicing spend happens on slipa_process, slipa_split, or slipa_create_video.

slipa_create_video

ChatGPT
Start generation on an already-uploaded project (same as slipa_process) and then WAIT, polling every few seconds, until the video reaches a terminal status — up to 10 minutes. Set verbatimNarration (SLP-310, below), same as slipa_process, to skip the narration-writing step and voice caller-supplied text unchanged instead. Use this when the caller needs the final video URL in one call rather than polling slipa_get_video itself. SPENDS MONEY: same as slipa_process — consumes one of the workspace's monthly video slots and real ElevenLabs budget — only call this with confirmSpend: true when the user has explicitly asked for a video to be generated. If generation is still running after 10 minutes this returns a timeout result with the projectId — keep polling with slipa_get_video, the pipeline is still running in the background. Resolve any named voice/music/brand/template/language option with slipa_list_voices / slipa_list_music / slipa_list_brands / slipa_list_templates / slipa_list_languages before calling this, same as slipa_process.

slipa_delete

ChatGPT
PERMANENTLY delete a walkthrough project: the project row, every retained variant (past generations), and every S3 object it owns (source video, narration, renders, posters, screenshots). There is no undo through this API. Free to call — does not spend a video slot — but destroys data, so only call this when the user has explicitly asked for a video to be deleted.

slipa_get_guide

ChatGPT
Get the written step-by-step guide Slipa generates alongside a video (the same content as the downloadable guide.md/guide.pdf) as markdown text, built from the project's storyboard. Built lazily on first call and cached after that. Requires the project to have a storyboard already (i.e. generation must have produced a script — call this AFTER slipa_process/slipa_split has progressed past scripting, not immediately after upload). Does not spend anything.

slipa_get_transcript

ChatGPT
Get the transcript for one walkthrough project as SRT subtitles, WebVTT captions, or plain text, built from the same script and caption timing the render pipeline itself uses. Reflects the requested generation: pass variantId to export a specific past generation (a retained variant from slipa_get_video's variants array) instead of the project's current text. Generations produced before transcript export existed retained no transcript and cannot be exported — only the current generation can be, for those. Does not spend anything.

slipa_get_video

ChatGPT
Get one walkthrough project by id: current status, generation progress (0-100), a human-readable status message, every retained variant (past generations from this same source recording), and a 24-hour signed download URL once status is "completed". Use this to check progress after slipa_process/slipa_split, or to fetch the final video link. Statuses: uploaded, transcribing, scripting, voicing, render_queued, cutting, rendering, uploading, completed, failed, preview_ready, expired — completed/failed/expired are terminal (generation has stopped either way). Does not spend anything.

slipa_list_brands

ChatGPT
List this WORKSPACE's own saved Brand kits: id (pass this exact number as brandId), name, and whether a logo is set — plus how many brands this plan allows and how many are already used. Call this BEFORE slipa_process, slipa_split, or slipa_create_video if the user asked to render with a specific named brand — brandId is data-driven and scoped to THIS workspace, never something to guess. Only lists brands actually created here — for the built-in Slipa look (used when brandId is omitted or explicit null), call slipa_list_templates instead. Does not spend anything.

slipa_list_languages

ChatGPT
List the narration OUTPUT languages Slipa supports: a code (pass this exact string as language), a display name, and whether that language has been end-to-end verified (script + TTS + burned-in captions all confirmed to render correctly) versus merely offered. Call this BEFORE slipa_process, slipa_split, slipa_create_video, or slipa_create_from_doc if the user wants narration in a language other than the project default — this is Slipa's headline capability (record in one language, publish narration in another, see language vs sourceLanguage), and language must be one of these exact strings. Does not spend anything.

slipa_list_music

ChatGPT
List the background music catalog: each track's slug (pass this exact string as musicTrack) and its human-readable label. Call this BEFORE slipa_process, slipa_split, slipa_create_video, or slipa_create_from_doc if the user asked for a particular mood or style of music — musicTrack is a slug from this catalog, not something to guess. Does not spend anything.

slipa_list_templates

ChatGPT
List Slipa's built-in visual templates (id, name, description) — the value to pass as template. These are the base look every workspace has before any of its own Brand kits are applied; call slipa_list_brands instead for a workspace's OWN saved brands (a different, workspace-scoped set, passed as brandId, not template). Call this before slipa_process etc. if the user asked for a specific visual look by description rather than by exact id. Does not spend anything.

slipa_list_videos

ChatGPT
List this workspace's walkthrough video projects, newest first, one page at a time (up to limit per call, default and max 50). Use this to find a project id before calling slipa_get_video, slipa_process, slipa_split, or slipa_create_video, or to answer "what videos do we have" / "show me recent generations". PAGINATION: the response always includes nextCursor and totalCount. If nextCursor is a non-null string, more videos exist beyond this page — call slipa_list_videos again with cursor set to that EXACT string to get the next page. If nextCursor is null, this was the LAST page — stop paging; a workspace with more than 50 videos is NOT unreachable past the first page, keep calling with the returned cursor until nextCursor is null. Never construct a cursor value yourself — only ever pass back a nextCursor this tool itself returned; a malformed cursor is rejected with a clean error rather than crashing or silently starting over. Does not spend anything, and does not include each project's full variants list (call slipa_get_video for one project's full detail, including its retained variants).

slipa_list_voices

ChatGPT
List every ElevenLabs voice this workspace can narrate with: id, name, and whatever labels ElevenLabs exposes for it (gender, accent, age, use case, description). Call this BEFORE slipa_process, slipa_split, slipa_create_video, or slipa_create_from_doc if the user asked for a specific KIND of voice ("a British male voice", "something energetic and young") — voiceId is data, not something you can guess or infer from a name, and passing a made-up id fails generation only after quota/render gates have already been checked. Does not spend anything.

slipa_plan_split

ChatGPT
Preview what slipa_split WOULD do, spending nothing: no project is created, no S3 object is copied, no render is kicked, and no quota is consumed. Returns, per group, the resolved segments, how many seconds of the source recording it keeps, and the projected output video length — plus totals (how many billable videos the real split would create, and the projected quota consumption). When options sets an explicit targetSeconds, also returns requestedSeconds and footageCapped per group — a group whose own kept-footage is shorter than the shared target gets footageCapped: true, so you can tell the user it will render shorter than requested BEFORE they spend anything on it. ALWAYS call this BEFORE slipa_split and show the user the plan — the same groups/options input, just without confirmSpend, since nothing here is billable.

slipa_process

ChatGPT
Start (or restart) the AI pipeline on an already-uploaded project: transcribe, write narration, voice it, cut, and render the final video. Set verbatimNarration (SLP-310, below) to skip the narration-writing step entirely and voice caller-supplied text unchanged instead. Requires an existing projectId — from slipa_create_upload_url (after its PUT has completed) or slipa_list_videos — for a project that is not already processing. SPENDS MONEY: consumes one of the workspace's monthly video-generation slots and real ElevenLabs text-to-speech budget — only call this with confirmSpend: true when the user has explicitly asked for a video to be generated. Returns immediately once generation has started; poll with slipa_get_video, or use slipa_create_video instead if the caller should wait for the result itself. If the user named a specific voice, music mood, brand, template or output language, look it up FIRST with slipa_list_voices / slipa_list_music / slipa_list_brands / slipa_list_templates / slipa_list_languages — every id/slug those options take is workspace data, not something to guess.

slipa_refine_script

ChatGPT
Rewrite a project's narration text (from its saved transcript — no re-transcription) while keeping the same beat structure, then re-voice and re-render. Use this for "make it punchier", "mention the pricing", or a tone/length change to the WORDING — for changing presentation (voice/brand/template) without touching wording, use slipa_regenerate instead. This tool ALWAYS re-renders (there is no free preview-only mode over MCP). Project must not already be processing. SPENDS MONEY: consumes one of the workspace's monthly video-generation slots and real ElevenLabs text-to-speech budget — only call this with confirmSpend: true when the user has explicitly asked for the narration to be rewritten. Poll the result with slipa_get_video.

slipa_regenerate

ChatGPT
Re-render a project from its SAVED script and beats — re-voice, re-cut, re-render — applying any changed render settings here (voice, music, template, brand, captions, resolution, aspect ratio) and optional per-beat edits (narration/sectionTitle/zoom on specific beat indices). Use this when the WORDS are right but the presentation should change (new voice, new brand, a tweaked beat), or to fix a handful of specific beats without touching the rest. To instead REWRITE the narration wording itself, use slipa_refine_script — that is the difference between the two. Set reScript: true (or any of instructions/tone/demoType/lengthPreset/targetSeconds/productName/language) to also re-run the AI storyboard from the saved transcript before rendering (still no re-transcription). Project must not already be processing. SPENDS MONEY: consumes one of the workspace's monthly video-generation slots and real ElevenLabs text-to-speech budget — only call this with confirmSpend: true when the user has explicitly asked for the video to be regenerated. Poll the result with slipa_get_video.

slipa_rename

ChatGPT
Rename a walkthrough project's title. Free — does not spend anything and does not affect the video itself. Also invalidates a cached written guide (slipa_get_guide will rebuild it with the new title on next call).

slipa_retry

ChatGPT
Retry a project whose generation FAILED, re-using the cached narration/render payload where possible (never re-paying ElevenLabs unless the original run never got that far, in which case this restarts the full pipeline instead). Only valid when the project status is "failed" and its retry count has not reached the per-project limit — call slipa_get_video first to check status. SPENDS MONEY: consumes a render/compute attempt against this project — only call this with confirmSpend: true when the user has explicitly asked to retry a failed video. Poll the result with slipa_get_video.

slipa_split

ChatGPT
Split ONE already-uploaded recording into several independent output videos in a single batch — e.g. one long capture covering three features becomes three separate videos, each with its own title and its own cut of the source footage. The project must not already be processing. SPENDS MONEY: creates one new billable project PER GROUP and consumes that many of the workspace's monthly video slots plus real ElevenLabs budget for each — only call this with confirmSpend: true when the user has explicitly asked for the recording to be split into multiple videos. Poll each resulting project with slipa_get_video. Same as slipa_process: use slipa_list_voices / slipa_list_music / slipa_list_brands / slipa_list_templates / slipa_list_languages to resolve any option the user named by description rather than by exact id, before calling this.

slipa_upload_limits

ChatGPT
Get this workspace's current plan limits for a new upload: the maximum single-video length in seconds (null means unlimited) and the maximum upload size in bytes. Check this BEFORE calling slipa_create_upload_url for a long recording, so the caller can warn the user up front instead of a file being rejected after it uploads. Does not spend anything.

Capabilities

Writes

Example Prompts

Click any prompt to copy it.

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/slipa"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/slipa)