NeuroCut
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
NeuroCut
An MCP server that gives an AI a multi-track video editor. It drives MLT — the engine behind Shotcut & Kdenlive — so the model lays down a timeline with terse tool calls, reviews cheap downscaled preview frames (never the whole video), iterates, and renders the real file to disk. Same philosophy as a good painting tool, for video.
Stateful project in server memory: video/audio tracks, clips, trims, speed, fades, per-clip & per-track filters, transforms (PiP/overlay), crossfades, undo/redo.
MLT under the hood → real compositing,
dynamictextcaptions,timewarpspeed,qtblend,avfilter.*(every FFmpeg filter). Projects save as.mltand open in Shotcut.Media & fonts from the internet —
add_media(url=…),add_text(font_google="Bebas Neue").Token-frugal: mutations return one line;
preview(at=…)andstoryboard(count=9)(a contact sheet of the whole edit) are the only things that cost image tokens, and they're downscaled;render_videocosts none.
Requirements
System packages (not pip): ffmpeg, ffprobe, and MLT with its Python
bindings + the melt CLI.
# Gentoo
sudo tee /etc/portage/package.use/neurocut-mlt <<< 'media-libs/mlt python ffmpeg xml qt6 rubberband opengl sdl gtk vorbis'
sudo emerge media-libs/mlt
# Debian/Ubuntu
sudo apt install melt libmlt++-dev python3-mlt ffmpegThen:
python -m venv --system-site-packages .venv # --system-site-packages so it sees mlt7
. .venv/bin/activate
pip install -e .
python -m neurocut # HTTP, 127.0.0.1:8766
NEUROCUT_TRANSPORT=stdio python -m neurocut # stdio for local Claude CodeRelated MCP server: ffmpeg-llm
Environment
var | default | meaning |
|
|
|
|
| HTTP bind |
| (unset) | if set, require |
|
| where |
|
| downloaded media & fonts |
|
| max bytes per fetched asset |
Connect as a Claude connector
./run.sh # starts server + a Cloudflare quick tunnel, prints URL + tokenIn Claude → Settings → Connectors → Add custom connector:
URL
https://<tunnel>/mcpAuthentication: None
Additional request headers:
X-Neurocut-Token=<token>
Tool map
new_project · project_info · list_projects · delete_project
add_media
add_track · update_track · remove_track
add_clip · add_color · add_text · move_clip · trim_clip · split_clip
· remove_clip · set_clip_speed · set_clip_fade · set_clip_gain
add_filter · add_transform · clear_filters · crossfade
preview · storyboard · render_video · save_mlt · get_download_link
· undo · redo
batch
render_video / save_mlt return a download_url (GET /dl/<token>/<name>,
opens in a browser with no auth) so the model can hand the finished file to the
user; get_download_link re-issues one for an earlier render. Set
NEUROCUT_PUBLIC_URL (run.sh does this from the tunnel) so the links are
absolute.
Time is in seconds. Track 0 is the bottom video layer; higher video tracks composite on top. Overlapping clips on one track are auto-laned, so text/PiP "just works".
Status / limits (v1)
Transitions: crossfade (alpha dissolve) + per-clip fades. Wipes/luma TODO.
add_transformusesqtblendrect; keyframed motion TODO.Audio track-level volume not yet wired (per-clip gain works).
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
This server cannot be deployed
Maintenance
Related MCP Connectors
AI video editor for agents and humans: timeline, captions, color, audio and generation as MCP tools.
AI video editor your agent can drive: plan scenes, edit a multi-track timeline, and render MP4s.
MCP server for Clipkit — gives AI agents a video toolbox via the Clipkit schema.
MCP server for Wan AI video generation
Related MCP Servers
- AlicenseAqualityCmaintenanceAn MCP server that lets AI agents visually perceive and edit video inside DaVinci Resolve Studio by rendering frames and manipulating the timeline.21MIT
- AlicenseNot gradedqualityDmaintenanceAn MCP server that exposes FFmpeg as a structured tool set for AI agents, enabling timeline-based video editing, preview, rendering, and analysis with an optional LLM autopilot.MIT
- AlicenseNot gradedqualityCmaintenanceA real video editor for AI agents, served over MCP, enabling journaled timeline editing, rendering via FFmpeg/MLT, and deterministic CLI operation.1MIT
- AlicenseNot gradedqualityCmaintenanceMCP server for OpenChatCut, enabling AI agents to create, edit, and export fully editable video projects through a real multitrack timeline and agent-native tools.1AGPL 3.0