Our take
- Activity
- 87
- Community
- 73
- Issues
- 56
- Pull requests
- 82
+50 stars in the last 8 days
- Merges more pull requests than 93% of the projects we track
- Releases more often than 89% of the projects we track
Overview
Memoh provides persistent workspaces for AI agents, rather than treating each interaction as a temporary chat session. Each agent has its own filesystem, desktop, browser, network, and long-term memory, and can remain online when the user's computer is off.
It is intended for people and teams who want to run multiple agents, connect them to messaging platforms, or host existing coding agents in dedicated environments. Memoh can be used through its cloud service or self-hosted.
Key Features
- Provides each agent an isolated workspace with a filesystem, desktop, browser, and network.
- Supports Memoh's built-in agent with user-provided API keys, as well as hosted Claude Code and Codex agents.
- Keeps agents running persistently, including when the user's laptop is closed.
- Supports interaction through Telegram, Discord, Lark, WeChat, and the Web UI.
- Provides long-term memory across sessions and platforms.
- Supports browser control, MCP tools, and scheduled tasks.
- Offers cloud access and self-hosted deployment options.
Use Cases
- Individuals can keep a personal assistant available across sessions and messaging platforms.
- Developers can run Claude Code or Codex in a persistent workspace instead of on their local computer.
- Teams can give members separate agents and workspaces on shared infrastructure.
- Operators can run multiple agents that need browser access, tools, or scheduled tasks.
Getting Started
For self-hosting, run the installer:
curl -fsSL https://memoh.sh | sh
See the README and deployment guide for setup options and configuration. Memoh Cloud is available at app.memoh.net.
Alternatives
- deer-flow: DeerFlow runs task-focused agent workflows with subagents, tools, memory, and sandboxes, rather than giving each agent an always-on desktop workspace.
- OpenHands: OpenHands centers on managing coding agents and recurring automations across backends, rather than persistent, isolated workspaces with desktops and browsers.
- lobehub: LobeHub focuses on building, coordinating, and scheduling agent workflows, rather than providing each agent with an isolated, always-on workspace.
| Project | Language | License | Stars | Status |
|---|---|---|---|---|
| Memoh | Go | AGPL-3.0 | 2.6k | Not checked yet |
| deer-flow | Python | MIT | 83.4k | Active |
| OpenHands | TypeScript | MIT | 89.9k | Active |
| lobehub | TypeScript | – | 83.1k | Active |
Considerations
- Self-hosting requires managing the deployment and configuration on your own infrastructure. Docker Compose is the documented full-stack option.
- The Compose deployment separates Server and Channel services and requires a private shared secret. A bare-metal setup can instead run as a single process when that secret is left empty.
- The AGPL-3.0 license should be reviewed before incorporating or modifying the software in a distributed service.
- The project is active and has 82 open issues, so check the repository for current status and known problems before relying on it in production.
- Running agents with API keys or connected accounts requires careful handling of credentials and workspace access.
Found this useful?
Share it with someone who would like Memoh.