English | 简体中文
Website · Getting started · Documentation · Report an issue
Molr is a self-hosted personal AI assistant that runs on your own device and connects to the models you choose. Use the desktop app or a browser on your computer or phone to delegate tasks through ongoing conversations, plan work, and review progress and deliverables in one place. You control each assistant's memory, file access permissions, and budget.
Warning
Project status: Molr is under active development. You may encounter instability, and future versions may introduce breaking changes.
Warning
Network access: Exposing the Molr service directly to the public internet is not currently recommended. Use it in a secure network environment, such as a trusted local network.
Molr provides desktop apps for Linux, macOS, and Windows, alongside the WebUI and Go Runtime. The desktop app can connect to an existing server or install and manage a local service on demand. Mobile users access the WebUI through a browser.
- Ongoing conversations and memory — Each assistant has its own context and memory, with tools to review, edit, and manage long-term memories.
- Plans and background execution — Schedule tasks, recurring work, and dependent handoffs with execution records. The browser can be closed while the service keeps running.
- Your models and budgets — Connect cloud or local models, assign models and budgets per assistant, and track usage and costs.
- Explicit resource permissions — Grant each assistant access to directories, tools, and compute resources. Linux, macOS, and Windows use their own execution backends and report the isolation and resource limits they support.
- Multiple devices and separate accounts — Continue in the desktop app or a browser connected to the same service, and use remote sandboxes to access other devices' files and execution resources. Families and small teams can allocate resources and quotas through user groups.
- Plugins and themes — Extend tools, Skills, themes, and standalone pages through plugins imported and authorized by an administrator.
Start by downloading the desktop app from the website, or use the command below to install a background service. Both options provide the latest stable release by default.
| Your use case | Recommended option |
|---|---|
| Start on your own computer with a graphical interface for installation and management | Desktop app: download it and create a local service in the app, or connect to an existing service |
| Run assistants on an always-on device or server for computers, phones, or multiple accounts | Background service (Runtime): run the installation command on the target Linux/macOS device, then connect with a browser or desktop app |
| Add another device's files or execution resources to an existing background service | Independent sandbox node (Sandbox): install it on the device providing the resources, then connect and authorize it from the existing service |
On a phone, open the running background service in a browser; no client installation is needed. Assistant tasks and data stay on the device running the background service. Keep that device awake, connected to the network, and running the service for tasks to continue.
Visit the website downloads and choose the desktop package for your system and architecture: .deb for Linux AMD64/ARM64, .dmg for macOS Apple Silicon/Intel, or .exe for Windows AMD64.
Open the app and choose to use this computer to create a local service, or connect to a remote server by entering an existing Runtime address. Full desktop packages include Runtime, WebUI, and the local sandbox, so no separate server package is needed; creating the local service still requires confirmation in the app. Closing the desktop window does not stop the running service. See desktop installation for the steps.
Run this command in a terminal on the Linux or macOS device that will run your assistants:
curl -fsSL https://molr.ai/install.sh | sh -s -- --mode serverThe installer detects the system and architecture, downloads and verifies the latest stable release, then guides you through service installation and administrator account setup. Windows users can create a local service through the desktop app.
The background service provides accounts, assistants, plans, WebUI, and a local sandbox. Connect to the reported service address with the desktop app or a browser; data stays on the server device. Linux uses systemd, with systemd 254+ and other sandbox dependencies required for controlled commands. macOS uses launchd and requires macOS 13 or later. See the installation guide for online installation, checksum verification, and maintenance.
To add another device's files or execution resources to an existing background service, run this command in a terminal on that Linux/macOS device:
curl -fsSL https://molr.ai/install.sh | sh -s -- --mode sandbox --name molr-nodeConnect and authorize the node from the background service's Resources page. A node provides no accounts, assistants, or WebUI. A single-device setup already includes a local sandbox and needs no additional node. See independent Sandbox services for connections and permissions.
Packages include neither model weights nor an inference server. Once the background service is ready, connect a local or cloud model, hire an assistant, and set its budget to start chatting.
- Getting started: sign in and start your first conversation.
- Installation and service management: install a specific version, choose a release source, or install offline. Historical releases and prerelease packages are available from GitHub Releases.
- Docker deployment: build the complete container image with WebUI.
- Building from source: build and deploy Molr yourself.
- Deployment and maintenance: HTTPS, sandbox dependencies, and backup and recovery.
- Multi-assistant group chats — Bring multiple assistants into one conversation, address participants with @mentions, and share group messages and deliverables.
See the full roadmap for scope, completion criteria, longer-term plans, and features under consideration.
Choose the user, administrator, or developer guide for your task.
| Guide | Contents |
|---|---|
| User documentation | Everyday use and troubleshooting |
| Administrator documentation | Installation, shared resources, members, and maintenance |
| Assistants, conversations, and plans | Ongoing conversations, scheduling, results, and revisions |
| Resources, permissions, and budgets | Model connections, file access, authorization, and costs |
| Plugins | Tools, Skills, themes, and page extensions |
| Architecture overview | Components, execution flow, and data storage |
| Linux command environment | Sandbox dependencies, limits, and capability checks |
| Developer documentation | Implementation, interfaces, branding, and contribution workflow |
Report problems through Issues, or submit a pull request to improve the code and documentation. Read the contribution guidelines first, and use a separate worktree for development, tests, and documentation changes.
The repository contains a Go Runtime, an Electron desktop app, three frontend applications, and shared packages:
| Directory | Purpose |
|---|---|
runtime/ |
Go service, agent execution, and sandbox |
apps/desktop/ |
Electron desktop app, connections, and local service management |
apps/webui/ |
Product interface connected to the real Runtime |
apps/website/ |
Product website and public documentation |
apps/ui-design/ |
Branding, components, and mock interactions |
packages/sdk/ |
Runtime client and public protocols |
packages/ui/ |
Shared components, design tokens, and brand assets |
packages/ui-host/ |
Theme resolution and isolation adapters |
Install dependencies in your task worktree, then start the application you need:
npm ci
npm run dev:desktop # Desktop development; local service and packaging prerequisites: apps/desktop/README.md
npm run dev # WebUI: 5191; requires a separate Runtime
VITE_SITE_MODE=live npm run dev:website # Website: 5189
npm run dev:design # UI Design: 5188; mock dataFrontends listen on 0.0.0.0 by default. Each worktree needs its own dependencies, data, and available ports. Set MOLR_RUNTIME_URL to connect the WebUI to that worktree's own Runtime. See each directory's README for configuration and validation commands, and the development guide for Runtime build and acceptance requirements.
Molr's own code and documentation are licensed under MIT. Third-party components retain their respective licenses. See Third-party notices and the licensing and distribution guide.