AgentIndex icon
AgentIndex
ToolsMCP ServersHOTCategoriesTrendingNewBest ListsCompareGuides
Submit Tool
ToolsMCP ServersHOTCategoriesTrendingNewBest ListsCompareGuides
Home/
Vision / Multimodal/
CozyClay
CozyClay logo

CozyClay

Active·★ 2.0k·AGPL-3.0·Updated 2026-11-18

CozyClay is open-source previs software that runs in a browser tab. Its premise, stated in the README, is that AI video costs money per take and a text prompt cannot hold a whole scene, so you decide where everyone stands, how they move and where the camera is in 3D first, where changes are cheap.

The workflow has four stages. Stage builds the set from primitives with rigged characters or capsule stand-ins. Move lets you describe an action; an agent writes a motion spec and a deterministic in-browser compiler bakes every frame, with QA for collisions, foot slide and hand contact. Shoot puts shots on a timeline with dolly rails and model-specific limits. Export produces a keyframe pack with first and last frame, greybox clip, camera and prompt, plus mp4, depth and normal passes and an OTIO cut list.

Install is npx cozyclay with Node.js 22.19+ and a Chromium browser; no GPU, Python or model download is needed. The scene can be driven from an in-app agent panel, an MCP server for Claude Code or Cursor, or a terminal CLI. It is licensed AGPL-3.0 and has anonymous telemetry that can be turned off.

This write-up is based on the project's README and public repository data. We have not hand-tested the tool. Data as of 2026-11-15.

License
AGPL-3.0
Latest release
v3.2.2 · 2026-11-07
Last commit
2026-11-15
Contributors
13
Forks
201
Open issues and PRs
7
First published
2026-07-28
npm package
cozyclay 3.2.2 · 850 downloads last week

Source: GitHub, npm and PyPI public APIs.

#previs#3d#ai-video#camera-control#storyboard#mcp#animation
$ Install
$ npx cozyclay
↗ Visit site★ GitHub
01

Features

01Stage sets from primitives with rigged characters or capsule stand-ins
02Agent-written motion specs compiled deterministically in the browser, with QA for collisions and foot slide
03Shot timeline with dolly rails that respects the target video model's limits
04Exports keyframe packs, mp4, depth/normal passes and OTIO cut lists
05Drivable from an in-app agent panel, an MCP server or a terminal CLI
02

Why choose it

+Plans framing and timing in 3D before you pay for any video generation.
+Exports a first frame, greybox clip and matching prompt so the video model follows the blocking.
+Runs with a single npx command and no GPU, Python or model download.
03

Trade-offs

!AGPL-3.0 licensing matters if you plan to embed or host it.
!Output quality still depends on the external AI video model you hand shots to.
!Requires a Chromium browser and Node.js 22.19 or newer.
04

Maintenance check

ActiveChecked: 2026-11-05
Last commit
2026-11-05
Commits, last 90 days
100+
Releases, last 90 days
30
Latest release
v3.2.1 (2026-10-28)
Open issues
14
Recent issues with at least one comment
2/22
Commits per week, last 12 weeks
Commits per week, last 12 weeks: 171, 187, 83, 78, 97, 33, 571, 130, 199, 229, 113, 37

Oldest week on the left; the latest four weeks are orange. Source: GitHub commit activity.

Status rule: archived on GitHub = Archived; last commit within 90 days = Active; 91 to 365 days = Slowing; older = Stale. All figures come from the GitHub API on the date shown.

Most discussed open issues

  • #1074 Agents cannot export a shot video: export.shotVideo always needs a click in the Studio1 comments · 2026-11-05
  • #1083 Measurement: authored-value count and first-compile qa.ok in compile output, bench and session stats (parent #1078)0 comments · 2026-11-05
  • #1082 Timing: tempo hints per key and `frames: "auto"` by default (parent #1078, item 7)0 comments · 2026-11-05
  • #1081 `look` target: gaze angle computed per frame instead of agent-keyed head angles (parent #1078, item 6)0 comments · 2026-11-05
  • #1080 Apply the walk's automatic rules to all keyed motion: knee pole, foot pitch/yaw, free-arm swing (parent #1078, items 3-5)0 comments · 2026-11-05

Issue titles are quoted from GitHub. Open the link to check whether a problem still applies to the version you use.

05

Compatibility

Chromium browsers
Web studio
Verified via docs
Claude Code / Cursor
MCP server
Verified via docs
06

Quick start

1
$ npx cozyclay
07

Use cases

↳Plan an AI video shot in 3D before paying for generation takes
↳Keep camera moves consistent across takes with a reference clip
↳Build storyboards and animatics for a short film
08

Alternatives

⚡Top VS:
VSCozyClay★2.0k→VSArcReel★6.6k→VSraea★2.7k→VSUnity-AI-ProBuilder★58→
CozyClay logo
CozyClay★ 2.0k
Open source previs software in the browser: block a scene, pose characters, author camera moves and cuts, then take the same shots to an AI video model.
VSCompare→
ArcReel logo
ArcReel★ 6.6k
AI Agent 驱动的开源可自部署视频工作台:将小说与剧本转为角色、场景、道具资产、分镜、视频和剪映草稿,支持跨镜头一致性、多供应商与费用追踪 | Self-hosted AI video workspace for stories, storyboards and short-form video production
VSCompare→
raea logo
raea★ 2.7k
Render Anything Engine: tell your AI what to make. Videos, motion graphics, 3D and games as React and TypeScript, rendered on your own GPU. Rust + Bevy, macOS, Linux and Windows.
VSCompare→
Unity-AI-ProBuilder logo
Unity-AI-ProBuilder★ 58
AI-powered 3D modeling tools for Unity ProBuilder. Enables AI assistants to create and manipulate editable meshes through natural language commands. Create primitive shapes, extrude faces, bevel edges, apply materials, merge objects, and perform advanced mesh operations like bridging and subdivision.
VSCompare→
huashu-design logo
huashu-design★ 29.7k
Huashu Design 3.0 · 在 agent 里打一句话,拿回能交付的设计:网页 / App 原型 / 可编辑 PPTX / 动画 / 信息图 · 六个候选挑三个方向 · 84 种带样张的风格库 · 越用越懂你的外观账本与设计体检 · HTML-native design skill for Claude Code, Codex, Cursor & any agent
VSCompare→
video-shotcraft logo
video-shotcraft★ 21.0k
AI video skill for Claude Code & Codex — cinematic, editable product promos with Remotion: 124 shot recipe cards, 150 motion previews, beat sync, sound design, Motion Workbench & JianYing (CapCut CN) export.
rea-video logo
rea-video★ 10.3k
Reverse-engineer any video, shot by shot: filmed, AI-generated or coded, and how to remake it. Breakdowns in the open VIDEO.md format. 逆向一切视频,逐个镜头看是拍的、AI 生成的还是代码做的,再告诉你怎么复刻;拆解结果按 VIDEO.md 格式输出。
autoclip logo
autoclip★ 10.7k
AutoClip|一个链接,一键出片。开源 AI 视频剪辑桌面工具,将播客、访谈、课程等长视频自动剪成短视频,生成字幕、封面和发布文案,适配抖音、小红书、TikTok、Reels 与 YouTube Shorts。Open-source AI video clipping & content repurposing.
VSCompare→
See all alternatives →

Comments

Log in to leave a comment

No comments yet. Be the first!

On this page
01Features02Why choose it03Trade-offs04Maintenance check05Compatibility06Quick start07Use cases08Alternatives
NomaDamas /
CozyClay
↗

Block a 3D scene and camera in the browser, then hand the shot to an AI video model

2.0k201AGPL-3.0
Last commitToday
StatusActive
CategoryVision / Multimodal
Contributors13
Latest releasev3.2.2
Links
Documentation↗Discussion↗Issues↗Releases↗
Cloud Infrastructure
Ad

Deploy on DigitalOcean

Host your MCP servers, Docker containers, and AI agents with $200 free credit over 60 days.

Claim $200 Free Credit
© 2026 AgentIndex.app|Built by a 10-year iOS Developer.
Best by use caseAboutContactAdvertisePrivacy PolicyTerms of UseQYSGitHubBuy me a coffee ☕

Browse by Category

MCP ServersCode AssistantWorkflow AutomationRAG / Knowledge BaseMulti-AgentBrowser AutomationLLM InfraDev ToolingObservability

Not affiliated with Anthropic, OpenAI or Microsoft.