Velorn

Open-source desktop video workstation with a local MCP server of 100+ tools and ComfyUI-backed generation.

AdaptableOperable ProductDesktop app Hybrid JavaScriptTypeScript MCP server License: GPL-3.0

An Electron editor that plans, generates, edits, captions, and exports inside one project. Generation runs through a local ComfyUI instance with custom API workflows; editing, captions, and export work without it. Agents connect to a loopback HTTP MCP server with preview-first write tools.

Use it when

I run ComfyUI locally and want a real timeline around it that my agent can review and edit
I make music videos or UGC ads from generated shots and want the whole project in one desktop app

Gotchas

GPL-3.0
All generation needs a separately installed local ComfyUI; heavy video workflows may want 24 GB+ VRAM
MCP binds to 127.0.0.1:19790 with no auth; any local process can call it
Imports, exports, and generated files are not undoable; the docs tell agents to preview and ask first

Last activity Aug 31, 2026. Fit: adaptable 1, operable 3.

Editions

Free · Free open source

DownloadDownload
Clonegit clone https://github.com/VelornLabs/velorn

For agents: JSON · Markdown

Adaptable: what is yours to change

Make it yours

Fork and modify1 thing to edit

GPL source with no documented env var or config file; settings live in the app. The intended extension point is a ComfyUI API workflow JSON with VELORN_* endpoint nodes, which is a feature rather than an adaptation. Changing the editor itself means forking the Electron app.

Env vars: none
Config files: ComfyUI API workflow JSON with VELORN_* endpoint nodes
Prompt files: none
Accounts you need: Comfy.org (partner-node cloud workflows only)Pexels (optional)

Source

github.com/VelornLabs/velorn
GPL-3.0 license
Last commit Aug 31, 2026
467 stars

Operable: what an agent can drive

Interface

NativeMCP serverpublished by the builderlocal

How an agent can drive an Operable Listing, in five levels: native (built for the agent as the primary interface), official (the Builder ships an agent interface), wrapped (a third party wraps the software's API), config (an agent can only drive it by editing files), gui (screen control only). An interface is official only when it is published by the same organization that ships the software. GUI-level Listings are allowed and sorted last.

Docs: github.com/VelornLabs/velorn/blob/main/docs/MCP.md
Tools: get_project get_timeline find_timeline_items analyze_timeline check_media_health add_timeline_markers create_project_checkpoint run_mcp_action_plan get_mcp_recipes

Reach

Partial coverage. 7 of 10 jobs reachable through the stated interface.

Job	Through the interface	Evidence
Review an edit: timelines, clips, frames, gaps, missing media	Reachable	docs/MCP.md: inspect timelines, frame contact sheets, analyze_timeline, check_media_health
Editorial changes: move, trim, split, retime, transitions, tracks	Reachable	docs/MCP.md 'Make safe editorial changes'; previewOnly then apply
Graphics: text, shapes, GLSL effects, keyframes, motion blur	Reachable	docs/MCP.md 'Build graphics and polish'
Queue generation jobs through ComfyUI	Reachable	docs/MCP.md 'Drive generation': queue approved batches, validate ComfyUI nodes
Import a community ComfyUI workflow and install its nodes	Reachable	README AI Agents: agent analyzes the graph, installs nodes after approval
Captions from timeline audio	Reachable	docs/MCP.md: caption actions exposed
Export H.264, social batches, FCPXML	Reachable	docs/MCP.md 'Deliver': H.264 delivery exports, social delivery batches, FCPXML
Music Video, UGC, and Business Ad guided creators	Undetermined	README Create tab; docs/MCP.md does not say whether Director Mode flows are callable
Search and import Pexels stock	Undetermined	README Stock tab; not listed among MCP workflows
Launch or restart ComfyUI	Not reachable	README: launcher and auto-start are Settings features

Code stack

Languages: JavaScriptTypeScript
Frameworks: ElectronReactffmpeg
Package manager: npm
Database: none
Hosting target: WindowsmacOSLinux
Built: Unknown (recorded, never ranked)

Model stack

Providers: local ComfyUIComfy.org partner nodes (cloud)LM Studio (optional assistant)Pexels (stock)
Models: LTX 2.3WAN 2.2Qwen Image EditNano Banana 2SeedanceKling
Keys you need: none
Prompts live in: none

Evidence and confidence

Field	Where it was seen	Confidence
formFactor	README Download: Windows installer, Mac, Linux AppImage; Electron	95%
executionModel	editing deterministic; generation through ComfyUI models and optional LM Studio assistant	85%
operability	README: 'a real editor for you, and 100+ MCP tools for your agent'; docs/MCP.md published by VelornLabs	90%
effort	no .env.example, no config marker; customization is in-app Settings and workflow JSON	70%
editions	GPL-3.0 LICENSE; velorn.ai says open source, available now; credits depend on providers	85%
active	GitHub pushedAt 2026-08-31	–
reach		75%

Open questions

Whether Director Mode creators and the Stock tab have MCP tools; count the tools in the source
Whether Velorn plans a paid cloud tier; velorn.ai mentions premium cloud workflows through providers only

Sources

velorn.ai sourcing:video-streaming:github
github.com/VelornLabs/velorn sourcing:video-streaming:github
github.com/VelornLabs/velorn/blob/main/docs/MCP.md analysis