NTM CLI
A tmux control panel that launches a swarm of coding agents, assigns tasks, enforces a safety policy and exposes REST and JSON.
- 464 ★ on the repository
- Last commit
Broad access
Why this level
- NTM starts and directs several agents with permission to run commands and edit files in one repository.
- Installation uses a network script and the local REST and WebSocket API opens control of the swarm.
Install
- Works in
Manual install
curl -fsSL "https://raw.githubusercontent.com/Dicklesworthstone/ntm/main/install.sh?$(date +%s)" | bash -s -- --easy-mode`tmux` must be installed; then `eval "$(ntm shell zsh)"` and `ntm deps -v`.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- NTM starts and directs several agents with permission to run commands and edit files in one repository.
- Installation uses a network script and the local REST and WebSocket API opens control of the swarm.
This is third-party code. Review the repository files before installing.
What it does
NTM (Named Tmux Manager) turns `tmux` into a local control plane for multi-agent development. One Go binary combines session launch, graph-aware work triage, a safety policy with approvals, mail between agents (Agent Mail), state capture and a local REST, SSE and WebSocket API. `ntm spawn api --cc=2 --cod=1 --agy=1` starts a mixed group of two Claude Code, one Codex and one Antigravity, and `ntm send api --cc "Map the auth layer and propose a refactor plan."` sends them a task. Sessions show in `ntm dashboard` and `ntm palette`. Work is picked with `ntm work triage` and `ntm work next`, and protection of dangerous operations comes through `ntm safety`, `ntm policy`, `ntm approve` and `ntm guards`. `ntm checkpoint`, `ntm timeline` and `ntm audit` give checkpoints, a timeline and an audit log. For automation there are `--robot-*` flags with JSON and an OpenAPI schema. `tmux` is required and the agents themselves install separately.
Who it is for. Developers who run several Claude Code and Codex agents in parallel on one repository and want to manage them from one place.
Good fit when
- Several agents must start at once over a repository and take tasks from a queue.
- Dangerous agent commands must pass through a policy and manual approvals.
- Agents need shared mail and file reservations so they do not overwrite each other.
- Swarm state must be kept in checkpoints with an audit log.
Not a fit when
- One agent in a regular terminal window is enough.
- There is no `tmux` and none is planned.
- Orchestration in the cloud with no local machine is needed.
Example request
Start an api session of two Claude Code and one Codex and send them a task to study the auth layer and propose a refactoring plan.Limitations
`tmux` is required and agents install separately: Claude Code, Codex, Antigravity CLI, Oh My Pi or Grok Build (Gemini CLI is supported as a legacy option). Many features rely on optional tools such as `br`, `bv`, Agent Mail, `cass` and `dcg`, which the README calls integration-heavy. `ntm deps -v` shows what is missing. Installation uses a network script. Strong agent permissions and parallel edits in one repository raise the cost of a mistake, so set up the policy in advance. The last commit in the repository is from November 2026.
How to disable. End the tmux sessions with agents and delete the `ntm` binary along with the `eval "$(ntm shell zsh)"` integration line from the shell rc file.
FAQ
How do I install NTM?
With `curl -fsSL "https://raw.githubusercontent.com/Dicklesworthstone/ntm/main/install.sh" | bash -s -- --easy-mode`. Then `eval "$(ntm shell zsh)"` enables shell integration and `ntm deps -v` checks dependencies.
Which agents are supported?
Claude Code, Codex, Antigravity CLI, Oh My Pi and Grok Build, with Gemini CLI kept as a legacy option. Flags in `ntm spawn` set the number of agents of each kind.
How do I guard against dangerous commands?
Through `ntm safety`, `ntm policy`, `ntm approve` and `ntm guards`: the policy blocks dangerous operations or requires a manual approval.
Is there an API for automation?
Yes: `ntm --robot-snapshot` prints JSON, `ntm serve` starts local REST, SSE and WebSocket and `ntm openapi generate` creates an OpenAPI schema.
How do I pick the next task?
With `ntm work triage` and `ntm work next`, which triage the queue by the work graph. `ntm assign` assigns tasks to agents.
Discussion
No questions or reviews yet.
Related
CLI toolsA CLI and terminal app that show which coding agents in your tmux sessions are working, waiting on you or done
A tmux-based TUI that runs Claude Code, Codex and other agents in one list: status, quick prompts, worktrees and diff review
ccmux tracks every coding agent in tmux: who is working, who waits for you, worktree spawning and answer hand-off.
A tuned tmux setup and the tmux-bridge CLI that lets agents read neighbouring panes, type into them and hand each other tasks.
An MCP server that lets agents in neighboring tmux panes read each other, hand off tasks and watch logs, no framework.
mngr runs coding agents in parallel locally, in Docker and on Modal and manages them like git: create, list, clone, message.
README in short
A single panel for a swarm of coding agents in tmux: launch, queue, safety policy, audit and an API in one Go binary.