No, not for this exact constraint. The README explicitly says Ubuntu Wayland host control is disabled, so you would need a cloud desktop or an isolated Local VM instead.

• Ubuntu 24.04 x64 has .deb and AppImage packages, but the Linux desktop build is still marked Beta and GNOME is the supported desktop environment.
• Bots can still use a cloud Linux desktop or an isolated Local VM, but those are not direct control of the Wayland host.
• Host control is documented for macOS and Ubuntu Xorg after explicit opt-in. Wayland remains disabled while issue #345 is being resolved.

If direct control of the local Wayland desktop is mandatory, this project does not currently satisfy the requirement. The README gives no resolution date for issue #345.

No, not directly. The README explicitly says that API keys are saved as plaintext in config.json; 0600 restricts Unix file access but does not satisfy a policy requiring secrets not to be stored in plaintext.

• The terminal mode supports Node 24+ and server deployment, but server deployment does not change the documented credential-storage behavior.
• openmausbot setup can reconfigure the model and access method, yet the README still states: “API keys are saved as plaintext, not encrypted.”
• Transcripts, events, and configuration are stored under ~/.openmausbot, so a VPS also requires controls for disk access, backups, and administrator access.

Using an already authenticated CLI may avoid this API-key path, but the README does not document how CLI credentials are stored and does not describe integration with an external secret manager.

Yes. It is designed for this exact setup: it organizes existing CLI agents into separate chat-like contacts instead of replacing them with a new model account.

• Every sidebar bot is a real agent with its own personality, model, thread memory, computer, and connected apps. The documented engines include locally installed claude, codex, and grok CLIs.
• macOS support covers both Apple silicon and Intel, and released desktop builds embed the harness server, so a separate server setup is not required.
• A bot can use a cloud Linux desktop, an isolated Local VM, or, after explicit opt-in, the macOS host computer. Important actions can require approval.

You still depend on the CLI login, subscription permissions, and network. The README does not specify whether context and tool behavior remain identical when moving a bot between different engines.

Yes, provided that pairing uses Tailscale, a protected HTTPS endpoint, or your own HTTPS reverse proxy. A phone cannot connect directly to a localhost URL.

• The README supports pairing a desktop client with another Windows, macOS, or Ubuntu host, and supports mobile browsers, native Android/iOS clients, and an installable phone web app.
• The terminal onboarding flow can configure phone access. Android uses a QR app link, while iOS accepts the app link or web link.
• Remote access must use an explicitly approved managed HTTPS endpoint protected by pairing, an existing Tailscale connection, or your own HTTPS reverse proxy. --local can ignore saved remote access for one launch.

This is suitable for remote access, but it does not automatically provide a secure public endpoint. The README does not specify the reverse proxy authentication implementation, device concurrency limits, or mobile support for every desktop feature.

Yes. The README explicitly provides terminal execution, CLI orchestration, and JSON output with exit codes for automation.

• You can install it globally with npm or run it through npx; the first launch guides AI access, model selection, and optional phone pairing.
• The project supports applying teams, running agent tasks, waiting for agents and collaborators to finish, and exporting results. JSON output and exit codes are intended for automation.
• The harness server owns agent processes, events, and sessions, while the UI and headless CLI share the workspace model.
• The documented requirements include Node 24 or newer and at least one installed and authenticated claude, codex, or grok CLI.

However, the README does not define CI retry, timeout, idempotency, or partial-completion semantics. Those details require validation outside the documented interface.

It depends. OpenMausBot is suitable for distributing role and workflow definitions, but the Markdown package should not be treated as a complete delivery of credentials, permissions, or runtime environments.

• The project supports Markdown team packages with YAML frontmatter describing bots, a Chief of Staff, channels, playbooks, connection lists, and recurring tasks.
• Imports are reviewable; connections are off by default and recurring tasks are paused by default. Packages do not carry credentials, sessions, permissions, or computer access, which improves portability and reviewability.
• The receiving user still needs an installed and authenticated claude, codex, or grok CLI, plus compatible model access and computer capabilities.

It is therefore best understood as a configuration and workflow blueprint, not a full copy of a running team. The README does not explain how differences in model permissions are detected or reported during import.