Overview
Newelle is a GPL-3.0 chat assistant for the Linux desktop, distributed as a Flatpak built on the GNOME runtime.
It is developed in the open at github.com/qwersyk/Newelle and listed on Flathub as "AI chatbot" under the developer name Qwersyk; the repository was created in June 2023. The latest release, 1.6.1, appeared on 2 November 2026 and the last commit on the master branch was on 6 November 2026, so development is active and no stale-project warning applies. PromptQuorum found no company, address or imprint, so it cannot say which country the project is based in.
According to the README, Newelle combines chat with several AI providers, downloadable local models, voice input and output, a real-time call mode, chat with documents, image generation, web search, a built-in file manager and terminal-command execution, plus an extension system and a mini window opened with a keyboard shortcut. The website adds Whisper dictation, Kokoro speech synthesis and hybrid vector-and-keyword document search; these are maker statements.
PromptQuorum has not run the app and has not checked speed, answer quality, voice accuracy or document-search quality.
How to Get It
On Linux you install Newelle from Flathub, which the project calls the recommended route; a software centre with Flathub enabled can do it.
The GitHub releases list no downloadable files, so Flathub is the packaged route. The README also describes a source build with install.sh or GNOME Builder, and development builds from GitHub Actions, for which it warns that translations will not work.
Route | How to get it |
|---|---|
| Flathub | flatpak install flathub io.github.qwersyk.Newelle |
| Website | Official site: docs, extensions, release notes |
| macOS (beta) | Newelle.app.zip, labelled Beta |
| Source | sh install.sh (GNOME SDK) or GNOME Builder |
No privacy-policy page was found on the website or in the repository. On 18 November 2026 PromptQuorum requested newelle.qsk.me/privacy, newelle.qsk.me/privacy-policy, newelle.qsk.me/terms, newelle.qsk.me/about, newelle.qsk.me/imprint and newelle.qsk.me/contact; each returned "404 Not Found". The file list of the GitHub repository contains no privacy, terms or legal file. PromptQuorum did not run the installer.
Requirements
Newelle needs a Linux desktop with Flatpak and Flathub set up; hardware needs depend on whether you use cloud providers or local models.
The list comes from the README and the Flathub listing. No minimum memory or GPU is published for local models; the 1.6.0 release notes say the model downloader estimates whether a model fits. PromptQuorum did not measure memory use or speed.
Item | Stated figure |
|---|---|
| System | Linux with Flatpak and Flathub |
| Architectures | x86_64 and aarch64 (Flathub) |
| Runtime | GNOME Platform 50 (Flathub) |
| Size on Flathub | About 336 MB download, 935 MB installed |
| Local models | Llama.cpp or Ollama; no minimum published |
| Build from source | Latest GNOME SDK, or GNOME Builder |
The size figures come from the Flathub summary on 18 November 2026 and change with each release. Local models are downloaded on top of that.
Features
Newelle's core is a chat assistant with a choice of models; voice, documents, tools and automation sit around it. All items below are maker claims from the README and release notes.
- Models: several AI providers and downloadable local models run with Llama.cpp or Ollama, plus a profile manager that keeps separate settings per profile.
- Voice: hands-free chat with many speech-to-text and text-to-speech models, wake-word support and a real-time call mode; the voice pill added in 1.5.2 is still experimental.
- Documents and memory: chat with your own documents, website reading with a # prefix, web search and long-term memory across chats, with a new explorable memory in 1.6.0.
- Tools: MCP servers added by pasting their JSON, skills from Claude and Codex, extensions, and an MCP server of Newelle's own tools for third parties.
- Automation: terminal commands suggested by the AI, scheduled tasks, subagent templates and interfaces that let you control Newelle from Telegram or an API.
- Images and files: image generation with Stable Diffusion or third-party services, vision input, a built-in file manager and editor, and permissions for which files Newelle may read or edit.
- Chat handling: multiple chats, workspaces, folders, branching from any message, editing, drafts, Markdown and LaTeX rendering, and a mini window opened with Ctrl+G.
Privacy and Data Flow
Where your prompts go depends on the model you choose: a cloud model sends them to that provider's servers, and a local model runs on your own computer. PromptQuorum did not inspect network traffic, so it cannot confirm that nothing else leaves the machine.
The Flathub listing shows that the packaged build can use the network and runs in a sandbox with display, audio and graphics access, without home-folder or session-bus access. The README says the Flathub version is restricted to its own data folder and describes two ways to widen that: home-folder access and permission to talk to the Flatpak host, so the AI can run commands outside the sandbox.
The README warns that these options reduce security by exposing your data and terminal, and that the maintainers cannot guarantee the privacy of your chat data or prevent risks from proprietary models. It also lists a file-permissions feature that sets which files Newelle may read or edit.
PromptQuorum found no privacy-policy page on the website or in the repository. The manifest in the GitHub repository lists home-folder and Flatpak-host access, while the Flathub metadata shows none; PromptQuorum did not check what each build grants. The code is GPL-3.0 and public, so the data flow can be audited.
Who Should Use It
Newelle fits Linux desktop users who want one free, open-source assistant for several providers and local models.
- Linux users who want a Flatpak chat app with local models, voice and document chat in one place.
- People who want to connect MCP servers, skills or their own extensions to a desktop assistant.
- Users who like to choose between cloud providers and local models for each chat or profile.
- Readers who prefer a project whose code and licence are public.
What We Could Not Verify
Newelle is a poor fit in some situations, and several details remain unverified.
- Not a fit: Windows users and anyone who needs a polished Mac or mobile app. The website offers Flathub and a macOS page labelled Beta; PromptQuorum found no Windows or mobile build.
- Unverified: speed, memory use, answer quality, voice accuracy, image quality and document-search quality. PromptQuorum has not run the app.
- Unverified: which AI providers and local models are supported today. PromptQuorum did not go through the documentation page by page.
- Unverified: whether the macOS beta is signed or notarized, and how it differs from the Linux build.
- Unverified: what the Flathub build and the repository manifest each grant, and how conversation data is handled when a cloud provider is used.
- Unverified: who runs the project, which country it is based in and how many people use it. No company, imprint or privacy policy was found, and PromptQuorum did not look for independent reviews.
- Unverified: that features listed on the website, such as Whisper dictation, Kokoro speech and hybrid document search, work as described.
Newelle vs Alternatives
Newelle competes with other desktop chat apps for local and cloud models; PromptQuorum has not compared it with them hands-on. The tools below are in the directory and are described from their own reviews.
Newelle
- Noted in sources:
- GPL-3.0; Linux (Flathub); free; voice and MCP
Articles about Newelle (2)
LM Studio
- Noted in sources:
- Proprietary, free; Mac/Windows/Linux; model browser
Articles about LM Studio (10)
- LM Studio Review: The GUI-First Local AI Desktop AppUpdated November 14, 2026
- Page Assist Review 2026: Browser Sidebar for Local AI ModelsUpdated November 20, 2026
- GenOffice Review: Open-Source AI Office Suite for Word, Excel and PowerPoint FilesUpdated November 20, 2026
- Ollama MCP Setup 2026: Connect Local Models to Files, Databases and APIsUpdated November 20, 2026
- Local Chat Apps & Assistants Compared (2026): Desktop, Mobile, Self-Hosted, Assistants and RoleplayUpdated November 20, 2026
- Local Inference Engines, Runtimes & Gateways Compared (2026): Run and Serve ModelsUpdated November 20, 2026
- Nanobrowser Review: Free Open-Source AI Browser Agent for Chrome and EdgeUpdated November 20, 2026
- Chatbox Review: Free, Open-Source AI Chat Client for Desktop, Web and MobileUpdated November 19, 2026
- Liquid Apollo Review: On-Device LFM2 Chat, Plus Optional CloudUpdated November 19, 2026
- llmfit Review: Find Which Local LLMs Fit Your HardwareUpdated November 19, 2026
+185 more not shown
Jan
- Noted in sources:
- Apache 2.0 (variant); Mac/Windows/Linux; offline chat
Articles about Jan (10)
- Jan Review: The Open-Source, Offline ChatGPT AlternativeUpdated September 6, 2026
- Local Chat Apps & Assistants Compared (2026): Desktop, Mobile, Self-Hosted, Assistants and RoleplayUpdated November 20, 2026
- Local Inference Engines, Runtimes & Gateways Compared (2026): Run and Serve ModelsUpdated November 20, 2026
- SoloMD Review: Free Markdown Notes App with AI Chat for Local Models and MCPUpdated November 18, 2026
- Local LLM Security and Privacy Checklist: 12 Steps to a Safe SetupUpdated November 15, 2026
- LM Studio Review: The GUI-First Local AI Desktop AppUpdated November 14, 2026
- Best Local LLM Apps for Mac in 2026: LM Studio, Ollama, Jan and 5 MoreUpdated November 13, 2026
- Atomic Chat Review: Local AI, Agents, and APIUpdated November 11, 2026
- Local LLM App Alternatives in 2026: Options to LM Studio, Ollama, Open WebUI and PocketPal AIUpdated November 10, 2026
- Best Local AI Apps for Low-End PCs in 2026 (8GB RAM, No GPU)Updated November 6, 2026
+43 more not shown
GPT4All
- Noted in sources:
- MIT; Mac/Windows/Linux; CPU-friendly local chat
Articles about GPT4All (10)
- GPT4All Review: Free Offline AI Chat App by Nomic AIUpdated October 24, 2026
- Local Chat Apps & Assistants Compared (2026): Desktop, Mobile, Self-Hosted, Assistants and RoleplayUpdated November 20, 2026
- Alpaca Review: The GNOME Desktop Chat Client for OllamaUpdated November 18, 2026
- Local LLM Security and Privacy Checklist: 12 Steps to a Safe SetupUpdated November 15, 2026
- LM Studio Review: The GUI-First Local AI Desktop AppUpdated November 14, 2026
- Atomic Chat Review: Local AI, Agents, and APIUpdated November 11, 2026
- Best Local AI Apps for Low-End PCs in 2026 (8GB RAM, No GPU)Updated November 6, 2026
- LM Studio vs Jan vs GPT4All: Which Local AI App Wins in 2026Updated November 3, 2026
- Best Local AI App for Non-Technical Users (2026)Updated October 28, 2026
- Locally Uncensored Review: Local Chat, Image, and Video in One AppUpdated October 24, 2026
+20 more not shown
AnythingLLM
- Noted in sources:
- MIT; Mac/Windows/Linux; document chat, vector store
Articles about AnythingLLM (10)
- AnythingLLM Review: Open-Source RAG and Document ChatUpdated September 11, 2026
- Local Chat Apps & Assistants Compared (2026): Desktop, Mobile, Self-Hosted, Assistants and RoleplayUpdated November 20, 2026
- Local Knowledge & Retrieval Tools Compared (2026): RAG, Document Chat, Vector Databases, Notes and SearchUpdated November 20, 2026
- Best RAG Tools for Business Documents 2026: Local & Private AI ComparedUpdated November 19, 2026
- Open WebUI in 2026: The Self-Hosted ChatGPT Interface for Ollama and Any AI ModelUpdated November 14, 2026
- Best Local LLM Apps for Mac in 2026: LM Studio, Ollama, Jan and 5 MoreUpdated November 13, 2026
- Local AI on Android in 2026: What Is Possible, What Is Not, and the Apps We CoverUpdated November 13, 2026
- Local LLM App Alternatives in 2026: Options to LM Studio, Ollama, Open WebUI and PocketPal AIUpdated November 10, 2026
- LM Studio vs Jan vs GPT4All: Which Local AI App Wins in 2026Updated November 3, 2026
- Best Local LLM Frontends & Chat UIs in 2026: Open WebUI, Enchanted UI, and MoreUpdated October 31, 2026
+37 more not shown
FAQ
Does Newelle run on Windows or Mac?
The website menu offers no Windows build. Its macOS page lists Newelle.app.zip, labelled Beta, version 1.3.7 from 21 April 2026, while the Linux version is 1.6.1. PromptQuorum did not test either.
Can Newelle run commands on my computer?
Yes, per the README it can execute commands the AI suggests. The README says the Flathub version is limited to its own data folder unless you grant home-folder or Flatpak-host access, which it says lowers security. The GitHub manifest lists host access while the Flathub metadata does not, and PromptQuorum did not check which applies.
Can I add my own tools or extensions?
Yes, per the maker: extensions add models and features, the documentation describes how to build an extension or a tool, and MCP servers can be added by pasting their JSON.
How do I open the mini window or voice pill?
Press Ctrl+G while Newelle is open, or start it with the --mini option and bind that command to a hotkey. The release notes say newelle --voice shows the experimental voice pill.
How often does Newelle update?
GitHub lists releases 1.4.5 on 19 June, 1.5.0 on 19 August, 1.6.0 on 15 October and 1.6.1 on 2 November 2026. Flathub installs update through the software centre.
Is Newelle free to use?
According to the README and the website, yes: Newelle is open source under GPL-3.0 and free to use. Cloud AI providers that you connect bill separately under their own terms, and local models are downloaded on top of the app. PromptQuorum did not check the pricing of individual providers.
Can Newelle run models locally?
Yes, per the README: it can download local models and run them with Llama.cpp or Ollama. No minimum memory or GPU is published; the 1.6.0 release notes say the model downloader estimates whether a model fits. PromptQuorum did not run a local model through it.
Where do my prompts go?
It depends on the model you choose. A cloud provider receives your prompts on its servers, while a local model runs on your own computer. The README says the maintainers cannot guarantee the privacy of your chat data. PromptQuorum found no privacy-policy page and did not inspect network traffic.
How do I install Newelle?
On Linux, the project names Flathub as the recommended route: flatpak install flathub io.github.qwersyk.Newelle, or a software centre with Flathub enabled. The GitHub releases list no downloadable files. The README also describes a source build with install.sh or GNOME Builder. PromptQuorum did not run the installer.
Does Newelle support voice?
According to the README, yes: hands-free chat with many speech-to-text and text-to-speech models, wake-word support and a real-time call mode. The website names Whisper dictation and Kokoro speech synthesis. The voice pill is still experimental, and PromptQuorum has not checked voice accuracy.
Can I chat with my own documents?
Per the README, yes: Newelle can chat with your own documents, and it can read a website when you use a # prefix. The website adds hybrid vector-and-keyword document search. These are maker statements, and PromptQuorum has not checked search quality.
Does Newelle support MCP?
Yes, per the README: MCP servers are added by pasting their JSON. It also lists skills from Claude and Codex, and the 1.6.0 notes add an MCP server of Newelle's own tools for third parties. PromptQuorum did not test any MCP server.
Who develops Newelle, and is it still maintained?
Flathub lists the developer as Qwersyk, and the GitHub repository was created in June 2023. The latest release, 1.6.1, appeared on 2 November 2026 and the last commit on the master branch was on 6 November 2026. PromptQuorum found no company, imprint or privacy policy, so it cannot say which country the project is based in.
Verdict
Newelle is worth a look if you use Linux and want a free, open-source assistant that works with several providers and local models and adds voice, document chat and tools.
On paper it covers a lot, and development is active: version 1.6.1 on 2 November 2026 and a commit on 6 November 2026. The Flathub listing shows a sandboxed build without home-folder access, while the GitHub manifest lists host access; PromptQuorum did not check what each build grants.
The trade-offs are real: no privacy policy, company or imprint was found, the macOS build is a beta, and PromptQuorum has not tested anything. Start with a small local model, check in your Flatpak permissions (for example with Flatseal) what your installed build actually grants, and confirm that voice, documents and tools work for your use before relying on them.
Sources
- Newelle website — features, install routes and macOS page, checked 18 November 2026.
- Release notes — notes for versions 1.6.1 and 1.6.0, checked 18 November 2026.
- macOS releases — the Beta build 1.3.7, checked 18 November 2026.
- Flathub listing — package, licence, versions and sandbox permissions, checked 18 November 2026.
- GitHub repository — licence, creation date and last commit, checked 18 November 2026.
- README — features, install and permission notes, checked 18 November 2026.
- GitHub releases — release dates, checked 18 November 2026.
- Documentation — extension, tool and MCP guides, checked 18 November 2026.
