SkillAgentSearch skills...

schneidi

Fast, light open-source video editor you can edit with an AI – chat in the app (Claude, OpenAI-compatible, Ollama), MCP server and CLI, offline subtitles. Linux AppImage/.deb.

Install / Use

claude mcp add OnkelTheorie -- npx -y github:OnkelTheorie/schneidi

If the server publishes to npm under a different name, use that package instead — check the repo README.

About this skill
🔌

MCP Server

Model Context Protocol server

Quality Score

84/100

Supported Platforms

Claude Code
Claude Desktop

Our assessment of schneidi

schneidi scores 84/100 on our quality scale, 2799th of 4,700 Development & Engineering skills we index.

Its MCP Server is 36 KB long, well organised into 16 sections with 7 code examples: a thorough specification that gives an agent plenty to work with.

It has 10 GitHub stars, so there is little community track record yet; judge it on its content.

Substance
30/30
Structure
20/20
Description
15/15
Adoption
4/20
Freshness
15/15

Maintenance, license and trust

  • The repository was last updated yesterday, so schneidi is actively maintained.
  • It is released under AGPL-3.0, a copyleft license: you can use it, but modified versions you distribute must carry the same license.
  • Its trust signals score 97/100, with no cautions. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.

Safety scan

No issues found

Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful.

AI review by kimi-k2.7-code on 2026-11-10. Automated pattern scan on 2026-11-09. It catches known dangerous patterns, not every risk — read a skill before letting an agent act on it.

schneidi compared with similar skills

All 4 of these similar skills score higher than schneidi; compare them before choosing.

SkillScoreStarsUpdatedFormat
schneidi (this skill)by OnkelTheorie84101d agoMCP Server
taste-skillby Leonxlnx100116.8k8d agoCLAUDE.md
Agent-Reachby Panniantong100112.3k9d agoCLAUDE.md
rufloby ruvnet10081.9ktodayCLAUDE.md
headroomby headroomlabs-ai10079.8ktodayCLAUDE.md

Frequently asked questions

How do I install schneidi?
Run claude mcp add OnkelTheorie -- npx -y github:OnkelTheorie/schneidi. The install tabs above show the steps for each supported agent.
Which AI agents does schneidi work with?
It is written for Claude Code and Claude Desktop, as a MCP Server file. Other agents that read the same format can often use it too.
Is schneidi safe to use?
Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful. It is AGPL-3.0-licensed and scores 97/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
Is schneidi still maintained?
The repository was last updated yesterday, so schneidi is actively maintained.

schneidi

A fast, light video editor that an AI assistant can edit with you – live, in the project you have open.

schneidi has five pages – Media, Edit, Comp, Color and Deliver – and fewer features than the big editors, but it starts in under a second, plays a demanding 1080p60 project smoothly on four CPU cores and runs even on a Raspberry Pi. It opens H.264/H.265 video with AAC sound on Linux out of the box – no codec packs, no paid version. The UI is available in English, German, Spanish, French, Polish and Russian.

Download

Download the latest release (Linux AppImage or Debian package). Windows: no download yet – build it yourself (tested less thoroughly than Linux); an installer is planned for 1.0.

Try the AI in two minutes: make the AppImage executable (chmod +x schneidi-*.AppImage) and start it, open a video, click ✧ AI at the top and pick a provider – a local model through Ollama needs no account (ollama pull qwen3, and start Ollama with OLLAMA_CONTEXT_LENGTH=32768 ollama serve: the tools and the editing guide do not fit its default context), Claude or an OpenAI-compatible service needs a key. Then type what you want: "remove the pauses", "cut it to the beat".

Status: early development (0.x). The project file format may still change between versions.

An AI assistant edits live in schneidi: keeps the speech, adds b-roll with dissolves, subtitles, a title, ducked music and a grade

<p align="center"> <a href="https://github.com/ZenITDigital"><img src="https://github.com/ZenITDigital.png?size=160" width="80" height="80" alt="ZenITDigital"></a><br> <b>💛 A huge thank you to <a href="https://github.com/ZenITDigital">@ZenITDigital</a></b><br> for sponsoring the Claude Max plan that schneidi is developed with. </p>

Why schneidi?

  • Edit together with an AI – any AI, any way. Tell it what to do in the AI chat next to the timeline (Claude, an OpenAI-compatible service or a local model), connect Claude Code or Claude Desktop in one step, or let an assistant work through the command line and MCP server, also on a server. It finds pauses, scene changes, loud moments and the beats of a song, transcribes, cuts, grades, mixes and renders – and you see what it does: changed clips light up, the AI log lists every step, each change is an undo step. Or review first and accept or discard the whole request.
  • From a recording to a short. A 9:16 short lives next to the 16:9 edit in the same project. Auto reframe keeps the subject in view (and a game's kill feed readable at the top), word-by-word captions come from the speech, and presets for YouTube Shorts, Instagram Reels and TikTok go with a check of the platform's length limit and loudness. An AI can do all of it in one call: schneidi-cli short turns game recordings into a finished short.
  • One template, many videos. Mark the clips, texts, song and logo of an edit as placeholders and it becomes a template: drop new clips onto it in the app, or let a script or an AI fill it from a table – one finished video per product, episode or match, all with your brand kit's logo, colors and fonts.
  • Layers, masks, keying and tracking – with nodes or as layers. The Comp page shows a clip's compositing as a simple layer stack or as a node graph: key a green screen, put text behind a person, blur a face that moves, stick a logo onto a surface. Ready recipes do the common jobs in one click, and an AI builds the same graphs through the command line.
  • A professional workflow on any computer. Pages for each step, ripple, roll, slip and slide trims, compound clips, multicam, color wheels and scopes – with the shortcuts you know from the big editors (a standard and a beginner preset, every key changeable), but light enough for an older laptop.
  • Everything stays on your computer. Speech recognition (whisper.cpp), text-based editing, sync by sound and rendering work offline – no account, no cloud, no telemetry, open source (AGPL-3.0-only). The only request schneidi makes by itself is a check for a new release on GitHub, at most once a day (off in File › Settings). The AI chat talks only to the provider you pick; with Ollama it stays on your computer too.
  • Long recordings of one event. Sync hours of recordings from several PCs or cameras by their sound and cut them in a multicam view that shows up to 16 angles at once.

Screenshots

Edit page while an AI assistant is connected: transcript, title inspector and timeline

The same edit as a 9:16 short in its own timeline: reframed picture, karaoke captions, picture shape in the Inspector Short – the 16:9 edit copied into a 9:16 timeline of the same project: reframed, with karaoke captions from the transcript

Comp page: a green screen recipe as a node graph – the keyed presenter in front of a title with a gradient, a sunset clip behind; the key's values in the Inspector Comp – the "text behind subject" recipe as nodes: key and refine the screen, a background and a title behind the person

File › New from Template: the five built-in templates as tiles with their preview, format, length and number of fields, then the start templates Templates – File › New from Template…: each template with its preview, format, length and fields; drag your clips onto the fields, or fill it from a table with schneidi-cli template fill

| | | |---|---| | Media page: Media Storage and Media Pool<br>Media – browse and skim folders, import into the Media Pool | Color page: a window darkens everything outside an ellipse, color wheels and RGB parade<br>Color – color wheels, windows, LUTs, scopes | | Deliver page: render settings and render queue<br>Deliver – presets and render queue | Multicam view: nine angles of one event side by side<br>Multicam – click an angle or press 1–9 to cut to it | | Comp page in the layer view: the same graph as a stack of layers with blend mode, mask and opacity<br>Layers – the same graph as a stack: blend mode, mask, opacity | A real 3D title: extruded, beveled golden letters turned in space, with the 3D section of the title Inspector<br>3D titles – real extruded letters with bevel, light and rotation |

Features

  • Multiple video and audio tracks: cut, trim, move, snapping, undo/redo
  • Multicam: a multicam clip from several recordings of one event, synced by sound, markers, In points or start; the viewer shows up to 16 angles at once (more on pages) – click or press 1–9 to cut, change angles later, flatten to normal clips
  • Timeline zoom and scroll change only on user input – nothing jumps when clips get shorter
  • Fully customizable keyboard shortcuts and mouse side buttons, two per function – presets (standard or beginner) and whole sets to save and load; panel sets for the Edit page, panel sizes remembered per page
  • Start templates: begin a project as a game recording, a podcast/interview or a vertical 9:16 video – or save your own setup (tracks, names, muted mix) as the start of every new project
  • Templates: a finished video from a template in one step – mark clips, titles, the song and the logo of any project as placeholders, then fill them by dragging files from the Media Pool onto them (File › New from Template… shows every template as a tile with preview, format, length and fields). A clip that is too short is trimmed, slowed, frozen or looped as the placeholder says, other shapes are reframed, long text gets smaller, the song is cut to length on its phrases. Five built in (a short with a hook, a gaming highlight on music, a 15 s product clip, a podcast clip, an intro/outro with logo); save your own and pass them on as a ZIP
  • Many videos from a table: one template, a CSV from Excel or a JSON list – one project and one video per row (schneidi-cli template fill product-15s --rows products.csv --out_dir out), e.g. a clip for every product
  • Brand kit (Workspace › Brand Kit): your logo with its corner and size, up to four colors, two fonts and an intro/outro clip – set once for all projects, changeable per project; titles take "brand color 1" instead of a fixed value and follow when the kit changes; logo and intro/outro go into the timeline only when you ask
  • Media Storage and Media Pool: look at a file and skim through it before importing, set In/Out and insert; filter the pool by flag, clip color and type; remove clips together with their uses in the timeline in one undo step
  • Watch folders: new files in a folder (your screen recorder's, for example) come into the Media Pool by themselves – only once they are finished; schneidi-cli watch for automatic workflows
  • Large viewer (one click hides the panels beside it) and full-screen viewer
  • Compositing (Comp page): a node graph on any clip, shown as layers (blend mode, mask, opacity per layer) or as nodes with links; nodes for clips, files, colors, gradients, text, merge, transform, crop, blur, grade and any effect; handles in the viewer; "Comp from Selection" turns stacked clips into one graph with the same picture
  • Masks and keying: rectangle, ellipse, polygon, bezier and gradient masks (soft edge, invert, combined), mattes from luma or alpha; a chroma and luma keyer with spill removal and a matte refiner (grow/shrink, soften, clip black and white)
  • Tracking: one point, two points (position, scale and rotation – for a sticker on a surface) or a face; masks and transforms follow the tracker
  • AI cutout (an optional extension): cuts out a person without a green screen, for text behind them or a new background
  • Recipes: green screen, text behind the subject, a blurred face that follows, picture-in-picture with a frame, a title with a mild glitch, a tracked sticker – one click in the Effects Library; save your own graph as a recipe and share it as a ZIP
  • Write your own effects: an FFmpeg filter chain or your own GLSL shader as an effect – written by you or an AI; code that comes with someone else's project or recipe runs only after you say you trust it
  • Titles (also an animated counter and a progress bar) with title looks – gradient or texture fill, a 2.5D edge, drop shadow, glow – and real 3D text (extruded, beveled, lit, turned in space), subtitles, fades, transitions (dissolves, dips, wipes, push and whip), volume, loudness normalization
  • Color correction, curves, LUTs, effects with keyframes in any order; a built-in Edit look preset (teal shadows, warm highlights, skin kept, a soft glow) in one click; copy and paste grades (=/-); the timeline filmstrip and the Media Pool show the grade; green screen with an eyedropper on a large picture of the clip
  • A format per timeline: a 9:16 short (or 1:1, 4:5) next to the 16:9 edit in one project; copy a timeline into another shape in one step; pasting between timelines of different shapes converts positions
  • Auto reframe: footage fills a frame of another shape and follows its subject – the view holds still, moves smoothly and jumps only at scene cuts; in game footage the kill feed stays readable as a cropped copy at the top. Or fit the whole picture in with a blurred copy around it, on one track; accents (zoom punch, flash, shake) on hi

Truncated for display — read the full file on GitHub.

Related Skills

View on GitHub
GitHub Stars10
CategoryDevelopment
Updated1d ago
Forks0

Languages

C++

Trust signals

97/100

From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.

1 info