Mjolnir CLI
Mjolnir runs Claude Code, Codex and other agent sessions in a terminal, in containers and over SSH, and keeps them running.
- 115 ★ on the repository
- Last commit
- GPL-3.0
Broad access
Why this level
- Agents run with access to your repositories, containers, SSH hosts and EC2, and sessions move between accounts.
- By default prompt and reply text goes to the Jev classifier through a public proxy.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/mjolnirDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Agents run with access to your repositories, containers, SSH hosts and EC2, and sessions move between accounts.
- By default prompt and reply text goes to the Jev classifier through a public proxy.
This is third-party code. Review the repository files before installing.
What it does
Mjolnir (the mj command) is one control plane for coding agents: Claude Code, Codex, Kimi Code, Grok Build, Muse Code, OpenCode and DSH run side by side on a laptop, in Docker, Podman and Apple containers, over SSH and on EC2. Each session has its own worker with a prompt queue and event journal, so the agent keeps working after the terminal closes or the daemon restarts. A session can move to another account, another agent or another machine, and verified checkpoints are kept until the old one is torn down. The dashboard shows remaining subscription quota across named profiles, bundles several repositories into one project and gives supported agents shared project memory. Before a push, independent reviewers, on a different provider when possible, check the finished code. Sessions are driven through mj new, mj prompt, mj message and mj wait, the HTTP API, or mj acp for stdio ACP clients. There is a terminal dashboard, a web viewer reachable from a phone over Tailscale and a desktop app.
Who it is for. Developers with several subscriptions, repositories and machines whose agent work should keep going while they are away from the computer.
Good fit when
- An agent works on a large task and the laptop is closed: the session must survive until morning.
- A session moves from a personal Codex account to a work one, or from Codex to Claude Code, without losing context.
- You need remaining quota across all profiles in one window.
- A task runs on EC2 or over SSH on a server and is watched from a phone.
- Before a push an independent reviewer from another provider checks the code.
Not a fit when
- You work with one agent in one terminal on one machine: the author says Mjolnir is not needed then.
- Prompt and reply text must not go to the external Jev classifier, which is on by default.
- You need a proprietary license: the project is distributed under GPL-3.0.
Example request
Create an mj session with a Codex profile on EC2, assign the task of fixing failing tests and tell me when it finishes.Limitations
By default Mjolnir sends recent prompt and reply text and help-search text to the hosted Jev classifier from TypeSafe through a public proxy. For sub-agents with adaptive effort it also sends the task name, instructions and model. The `[jev] enabled = false` setting in `config.toml` turns this off. Linux releases need glibc 2.28 or newer on x86-64 or ARM64, and Windows works through WSL2. Muse cannot yet receive the injected shared-memory tools. The license is GPL-3.0-only.
How to disable. Stop the daemon and remove the package with npm uninstall -g @brokkai/mjolnir if you installed through npm. Disable agent profiles in settings rather than deleting them.
FAQ
How do I install Mjolnir?
On Linux or macOS run `curl -fsSL https://raw.githubusercontent.com/BrokkAi/mjolnir/master/install.sh | bash` or `npm install -g @brokkai/mjolnir`. Windows needs WSL2.
Where do I start after installing?
Run `mj` in the project directory: on first launch it finds installed agents and the repository. Create a session with `ctrl+b c`.
What does mj simple do?
`mj simple` is a minimal tmux-friendly mode: it uses the default agent, starts the session locally in an isolated checkout and disables automatic review.
How do I check prerequisites?
Run `mj doctor`: it checks the environment and says how to fix anything missing.
How do I stop sending data to Jev?
Set `[jev] enabled = false` in `config.toml` and Jev requests stop.
Discussion
No questions or reviews yet.
Related
CLI toolsA local control panel for coding agents: it runs Claude Code and Codex, keeps a task queue and merges verified work into main.
A tmux-based TUI that runs Claude Code, Codex and other agents in one list: status, quick prompts, worktrees and diff review
A web console for developers running several agents across projects: sessions survive crashes and replies come from a phone.
A CLI with a TUI that finds Claude Code, Codex and Gemini CLI sessions and continues them in another agent as native ones
PromptOps keeps Claude Code, Codex and Gemini CLI side by side by project and gives a lead agent MCP to launch helpers
A terminal for running many coding agents: Claude Code, Codex and OpenCode sessions on a laptop and over SSH in one window.
README in short
The project was built by the Brokk team for their own use and is open source under GPL-3.0-only. Documentation is on mjolnir.brokk.ai: overview, targets, profiles, sessions, durability, review and security. The previous generation, Mjolnir 1.x, is available in releases.