iArt.ai
AI 모션 그래픽 소프트웨어, 누구나 사용할 수 있는 After Effects 대안, 텍스트 프롬프트로 몇 분 안에 애니메이션·설명 영상 제작
Arynwood MCP는 Linux용 로컬 우선 AI 작업공간입니다
Arynwood MCP is a local-first AI workspace for Linux: Ollama chat, knowledge search with Qdrant, a design canvas and MCP tool integrations. Try the browser demo or download the v0.4.3 alpha.
미디어TypeScript · PythonGitHub @Arynwood-Technology온라인운영자 미확인
써 보셨어요? 한 번만 눌러 주세요. 로그인은 필요 없어요.
Local-first AI desktop workspace for Linux: Ollama chat, knowledge search, design canvas and MCP tool integrations (incl. Kdenlive) on your own hardware. Alpha, AppImage + .deb.
저장소 README에서 가져옴
README 전체를 불러오는 중입니다. 아래는 저장된 발췌입니다.
Local-first AI creative studio. One app for multi-persona chat, GPU generation tools, video editing (Kdenlive automation), audio production, design, and social publishing - all running on your own hardware.README 전문 보기 ↗A personal AI and creative workspace for your desktop. Work with local models, your documents, a design canvas, audio and video tools, and supervised Model Context Protocol (MCP) integrations. Connect your own Arynwood Grove for shared calendars, files and community spaces.
Website · Downloads · Documentation · Browser demo
Arynwood MCP is an open-source desktop alpha, for one owner on one machine.
Linux x86_64 AppImage and .deb packages and a Windows x64 installer are available.
0.5.4 is published, with signed checksums and verified build provenance.
Download it from Releases.
Version 0.5.4 brings security fixes, stricter desktop browser policy, PCM microphone recording, a first-play WebAudio audition fix and updated desktop dependencies. It also adds hashed Python build dependencies and bounded IRC property fuzzing. Release notes · Verification record.
Main requires pull requests and passing backend, frontend, API-reference and Windows checks. CodeQL and OpenSSF Scorecard run alongside the test workflows. Arynwood MCP has earned the OpenSSF Best Practices Passing badge through its public project self-assessment. The badge reports evidence against published criteria; it is not a security certification. Dependency advisories and maintenance findings are tracked separately in the Scorecard follow-up.
| Work | What MCP provides |
|---|---|
| Chat and research | Local or configured remote model endpoints, personas, document retrieval and owner-controlled memory |
| Creative work | Design canvas, recording and audio workflows, captions and video editing integrations |
| Supervised tools | MCP server connections, including Kdenlive editing through Cutroom's tool server |
| Community | Your Grove's calendars, tasks, files, messages and website previews inside the app |
| Models and services | Endpoint configuration, service status and separately installed creative sidecars |
Models, Qdrant and creative sidecars are installed separately. Some workflows need additional hardware or a source checkout. Read the feature guide, known limitations and packaged-build exclusions before choosing a workflow. The gateway is experimental, owner-only and off by default.
The browser demo uses the real interface with clearly marked fixture data and scripted conversations. It runs no backend or live model.
Download the latest published packages from GitHub Releases.
.deb with sudo apt install ./arynwood-mcp_*.deb.Linux packages target glibc 2.35 or newer. Models and supporting services are separate; a GPU is optional for basic chat, while generation tools have their own requirements. Platform requirements · Installation and first run.
Verify downloads before running them. Published 0.5.3 has build checksums but
no provenance attestations or release-key signature. Release 0.5.4 includes
per-package GitHub attestations and owner-signed SHA256SUMS with its detached
SHA256SUMS.asc signature. Follow the instructions for the version you download.
Download verification.
Requires Python 3.11+, Node.js 20.19+ or 22.12+, npm and a separately configured model service.
git clone https://github.com/Arynwood-Technology/ArynwoodMCP.git
cd ArynwoodMCP
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
npm --prefix frontend ci
cp .env.example .env
./start.sh
Configure the copied environment file for your setup. The interface opens at http://localhost:5180 and the API at http://localhost:8010. Knowledge search also needs Qdrant and an embedding model. Source setup · API reference.
Your default model inference, conversations and memory run on your hardware. Network features contact outside services when used: web search, URL learning, downloads, publishing, deployment, social services, IRC and configured remote endpoints. There is no enforced offline mode or OS sandbox.
Destructive and publishing MCP calls require approval for the exact operation. Direct owner controls have separate limits, and this is not a hardened multi-user service. Read SECURITY.md and scope. Report vulnerabilities privately through GitHub security advisories.
Arynwood Technology maintains the scope, architecture, review, tests and releases. Contributing · How we build · Project history.
Code is licensed under GNU AGPL-3.0-only. Releases through 0.5.3 used an earlier source-available license. The Arynwood tree logo and app icons are separate artwork, all rights reserved; the code license does not grant rights to that branding.