Skip to main content
PromptQuorumBuilt for humans. Structured for AI.
Home/Power Local LLM
Overview & Reference

Arynwood MCP Review: A Local AI Workspace in Alpha

·8 min readBy Hans Kuepper · Founder of PromptQuorum · Discovery engine for open-weight & open-source AI

Arynwood MCP (version 0.5.3, released 23 October 2026) is a desktop alpha for Linux and Windows that, per its website, runs chat, document search and tool integrations on your own machine through Ollama. Every statement in this review comes from the project's own pages; PromptQuorum has not installed or tested it.

Key Takeaways

  • Maintenance: version 0.5.3 was published on 23 October 2026, the latest upstream update at the time of checking; the repository was created on 10 September 2026, so no stale-project warning applies but the project is about six weeks old.
  • Licence: source-available, not open source, and limited to personal or internal use.
  • Installers: unsigned on Windows, and the whole app is labelled alpha.
  • Data: the website says there is no telemetry, yet several features send data out and no offline mode is enforced.
  • Basis: the website, README, licence file and release notes, read on 24 October 2026, with no hands-on testing.

📍 In One Sentence

Arynwood MCP is a Linux and Windows desktop alpha that chats with local Ollama models and adds document search, memory, a design canvas and MCP tool integrations.

💬 In Plain Terms

Think of it as one app window around a local chat model, a document library and a toolbox, installed from a downloaded package rather than a store.

📌Note: PromptQuorum has not installed, tested or benchmarked Arynwood MCP. Statements below about features, privacy and security are the project's own.

Read in:
🇺🇸en🇩🇪de🇫🇷fr🇯🇵ja🇨🇳zh🇪🇸es🇧🇷pt🇸🇦ar🇰🇷ko

Arynwood MCP is a desktop AI workspace that, per its website and README, chats with local models through Ollama and adds document search, memory, a design canvas and MCP tool integrations. It is a version 0.5.3 alpha for Linux and Windows under a source-available licence. This review reflects the website, README, licence file and release notes as checked on 24 October 2026.

Arynwood MCP Review: A Local AI Workspace in Alpha

What Is Arynwood MCP?

Arynwood MCP is a local-first personal AI and creative workspace from Arynwood Technology, built by Lorelei Noble, that sends chats to models running in Ollama on your own hardware. "MCP" stands for Model Context Protocol, which the app uses to connect tools; the README and website describe a desktop workspace, not a plug-in for other programs.

PromptQuorum found no other product called Arynwood MCP. The same maker's Arynwood Grove is a separate community server that the app can connect to, and the website's browser demo shows the interface with scripted conversations and no real AI behind it.

Get It

Arynwood MCP is downloaded from the project's GitHub Releases page, with one package per system; Ollama is installed separately.

System
Download
Linux x86_64 (AppImage)arynwood-mcp_0.5.3_amd64.AppImage
Debian / Ubuntu (.deb)arynwood-mcp_0.5.3_amd64.deb
Windows x64 (unsigned alpha)arynwood-mcp_0.5.3_x64-setup.exe
ChecksumsSHA256SUMS.txt (Linux), SHA256SUMS-windows.txt
macOSNot supported
Websitearynwood.com/mcp
PrivacyWhat leaves your machine (no separate privacy policy)
Security briefarynwood.com/mcp-security
Source codegithub.com/Arynwood-Technology/ArynwoodMCP

Version reviewed: 0.5.3 (desktop alpha), published on 23 October 2026 per the GitHub release page, which is the latest upstream update at the time of checking. The Windows installer is unsigned, so Microsoft Defender SmartScreen may show "Windows protected your PC"; the maker's guide says to compare the SHA-256 checksum first and then choose More info and Run anyway. The AppImage runs after chmod +x, and the .deb installs with sudo apt install ./file.deb. The website has no separate privacy policy; its "What leaves your machine" section is the privacy statement.

What You Need

Arynwood MCP needs Linux x86_64 or Windows x64 with Ollama installed separately; the table gives the memory, disk and graphics needs. The documentation adds that 8 GB can work with a smaller model or a remote endpoint.

Item
Requirement per the docs
Linuxx86_64; AppImage needs glibc 2.35 or newer
Windowsx64 alpha, per-user installer with WebView2
macOSNot supported
Memory and disk16 GB RAM; 20 GB free plus models
Graphics cardOptional for chat; NVIDIA for image/music/video
Other servicesOllama required; Qdrant for knowledge search

Without an NVIDIA card the app switches to a CPU mode that disables the GPU tools and tunes chat for the processor; the documentation warns that an 8B model can take minutes per reply on a small server. Ollama, Qdrant and the GPU tools are separate services, not bundled in the installers.

Pricing and Licence

The website lists no price for the app, and the packages are published for download without a fee; the Arynwood MCP Source-Available License, however, is not an open-source licence, per the website, and sets limits. Setup services from the maker are quoted on request.

  • Allowed. Downloading and running unmodified builds for personal or internal evaluation, and viewing and modifying the source for personal or internal use.
  • Not allowed without written permission. Redistributing, selling or sublicensing the software or any derivative, and offering it as part of a hosted or managed service.
  • Alpha warranty. The licence text provides the software as-is and warns of breaking changes, data loss and incomplete features; third-party components keep their own licences.
  • The maker's own caveat. The licence file calls itself "a template starting point, not reviewed by a lawyer" and tells readers to have counsel review it before relying on it for a public release.

How to Get Started

The steps follow the maker's installation and Windows guides; PromptQuorum has not run them.

  1. 1
    Compare your computer with the requirements
    Why it matters: Check memory, free disk and operating system against the table above before downloading.
  2. 2
    Install Ollama and pull a model
    Why it matters: Ollama is not bundled. Install it from its own site, then pull the models the personas use, or enter the exact name of a small model on a CPU-only computer.
  3. 3
    Download the package and check the checksum
    Why it matters: Save the package for your system and the matching checksum file into one folder and compare the SHA-256 value before you run anything.
  4. 4
    Install or start it
    Why it matters: On Linux make the AppImage executable and start it, or install the .deb with apt. On Windows run the installer, which installs for your user account only; choose More info and Run anyway if SmartScreen warns.
  5. 5
    Open the app and connect the services
    Why it matters: The app tells you if it cannot reach Ollama. Knowledge search additionally needs Qdrant, which is also installed separately.

Features Confirmed by the Sources

Every item below comes from the website, README or the maker's documentation.

  • Chat. Multi-persona chat streamed from Ollama, with model management and registration of OpenAI-compatible servers.
  • Knowledge and memory. Teach it from a URL, file or text; semantic search is added to chat through Qdrant and an Ollama embedding model, and the chat keeps memory and accepts file uploads.
  • MCP tools. Connects MCP tool servers, including Kdenlive editing through Cutroom; the security brief says one approval authorizes one exact tool call, once.
  • Design Center. A canvas design tool whose image generation uses a separately installed Stable Diffusion (A1111) service.
  • Audio and video. Optional studios that rely on separately installed sidecar tools; LoRA training, script-based GPU tools and the project file browser need a source checkout, not the packaged build.
  • Community. An optional page that connects to an Arynwood Grove server (0.7.1, alpha); messages are kept on that server and are not end-to-end encrypted.
  • Gateway. An experimental headless gateway for owner-only remote control, off by default and run from a source checkout.

Privacy and Network Use

The website says there is no telemetry, and that chats, documents and memory stay on your disk. It adds that the app checks GitHub for a newer stable release at startup and every six hours while open, sending no credentials or workspace content, though GitHub receives normal connection data such as your IP address.

Other data leaves the machine only through features you use, per the website: web search, learning from a URL, downloads, social publishing, deploys, IRC, a Grove on a server, or a remote model server you add. The security brief states there is no operating-system sandbox or enforced offline mode and that the stated scope is one owner on one machine; local chat, document search and memory can work offline once models are downloaded.

📌Note: These are the maker's statements. PromptQuorum has not audited the code or the network traffic, and the security brief itself says its tests are not a penetration test or a security certification.

Who Should Use It

  • Linux or Windows users who want an Ollama workspace in one app.
  • People comfortable with alpha software, an unsigned installer and a personal-or-internal-use licence.
  • Kdenlive editors who want to try tool-calling through Cutroom.

What We Could Not Verify

  • Hands-on behaviour. PromptQuorum did not run Arynwood MCP, so feature descriptions and the quality of local answers are the project's statements.
  • Windows install. The release notes list a clean Windows machine install and posting to a real social account as not run for 0.5.3, and the Windows guide says downloads and external integrations are not yet manually validated.
  • Privacy and security claims. The no-telemetry, update-check and data-flow statements were not checked against the code or network traffic.
  • Track record. Versions 0.5.2 and 0.5.3 both appeared on 23 October 2026, in a repository created on 10 September 2026; there is no long release history to judge by.

Competitors and Alternatives

Msty

Platforms:
Windows, macOS, Linux
Price / licence:
Freemium / closed source
Key difference:
Consumer-style chat interface with several model providers
Articles about Msty (6)

Also mentioned in:

+6 more not shown

Competitor details change often; confirm each app's current price, licence and platforms on its own listing.

Compare in directory

Frequently Asked Questions

Does it work without internet?

Local chat, document search and memory can work offline once the models are downloaded, but there is no offline switch, and the update check and web-based features still reach for the network.

Does it run on a Mac?

No, macOS is unsupported according to the maker's platform guide. The browser demo on the project's site shows the interface without any install.

Can it use a model server on another computer?

The Servers page registers Ollama or OpenAI-compatible endpoints, and the documentation says an endpoint can run chat and image generation on another computer instead.

How are updates installed?

Versions from 0.4.8 show an update notice; installing is manual by running the newer installer or package, and earlier versions need one manual install first.

Can I verify the Windows installer?

The maker publishes SHA-256 checksums for each package and says builds from 0.5.4 on can also be checked with GitHub attestations. PromptQuorum has not verified either.

Verdict

Arynwood MCP is a Linux and Windows desktop alpha that, per its website, brings Ollama chat, document search, memory and MCP tool integrations into one local-first app, with public installers and a release published the day before this check. Against that, the licence is source-available and limited to personal or internal use, the Windows build is unsigned, macOS is unsupported, no offline mode or sandbox is enforced and the repository is about six weeks old. It suits people comfortable with alpha software who accept those limits; readers who want an open-source licence or a longer track record can compare Jan or AnythingLLM.

Sources

← Back to Power Local LLM

MCP