Why mngr
Simple
One command launches an agent locally or on Modal with sensible defaults.
Cost-transparent
Agents auto-shutdown when idle and you pay only for inference and compute.
Secure
Isolated SSH keys, allowlisted network access, and full control over the container.
Composable
Shared hosts, snapshot and fork agent state, direct exec, push/pull/pair.
Observable
Transcripts, direct SSH, programmatic messaging.
Easy to learn
Use "mngr ask" to answer usage questions without leaving the terminal.
All your agents in one place
mngr list displays all of your agents with their status, host information, and other metadata. Supports filtering, sorting, and multiple output formats.
Starts up in ≤ 2 seconds
mngr create sets up an agent's working directory, optionally provisions a new host or uses an existing one) runs the specified agent process, and connects to it by default.
Agents that work like git
Think of mngr as "git for agents": just like git allows you to commit/push/pull/fork/clone versions of code, mngr allows you to create/destroy/list/clone/message agents.
lets you do things like:
— for each open GitHub issue, create a PR
— for each flaky test in the past week, fix it
— for each rule in style guide, scan codebase & fix all instances
The future of agentic systems is scaling out, not up...this is the way
I really like the Unix philosophy of mngr, mainly using background no connect agents which works a lot better than native Claude code subagents
Frequently asked questions
What is mngr?
A Unix-style CLI for running coding agents in parallel. One command launches an agent on your laptop, a Modal sandbox, or a Docker container. Same commands whether you run one agent or a hundred.
Which agents and hosts does it support?
Claude, Codex, OpenCode, or any agent you can launch from a shell. Hosts can be local, Modal, Docker, or anything else you can SSH into. Plugins let you add your own agent types, providers, and commands.
How fast is it?
It’s fast. Depending on your setup, agents can start and mngr list returns within a few seconds. There's no managed service in the middle. Just SSH, git, and tmux.
What does it cost?
The CLI is free and MIT-licensed. You pay for inference and the compute on whichever host you pick. Remote hosts auto-shutdown when agents go idle, so a Modal run typically costs a fraction of a cent in compute beyond inference.
How are agents isolated?
Each agent gets its own container, accessed over a unique local SSH key. You can block all outgoing internet with -b offline, restrict traffic to an IP allowlist, or layer on the provider's native isolation. If an agent does something reckless, the blast radius stops at the container.
How do I install it?
One line:
curl -fsSL https://raw.githubusercontent.com/imbue-ai/mngr/main/scripts/install.sh | bash