Switching to NeuroSquad
Used another app to run several coding agents at once? Most of what you know carries over — only the names change. This page maps the usual pieces to where they live in NeuroSquad, then rebuilds a setup step by step.
NeuroSquad is free: no trial, no credits, no limit on agents or workspaces. You pay only for the models — your own CLI subscriptions, or OpenRouter credits if you use them. Agents, a model per agent, the cost of each agent, your phone and on-device dictation are all in the same app.
What maps to what
| You had | In NeuroSquad |
|---|---|
| A project | A workspace — one project folder. It can be on this computer, in WSL or on an SSH host. See Your first workspace. |
| Agent panes or terminals | Agent cards, each with a live terminal, on an infinite canvas or in grid mode — CtrlShiftG (CmdShiftG on a Mac) switches between them without restarting anything. |
| A list of supported CLIs | 22 agent CLIs — Claude Code, Codex CLI, OpenCode and the rest. The CLI you already use runs as it is, on your own sign-in. |
| A dashboard of running agents | The Squad board and the statuses on every card: working, needs your input, finished. See Finished / needs your input. |
| A git worktree per agent | Isolate in git worktree before adding an agent, plus a setup command and files to copy into each new worktree. See Isolated worktrees. |
| Picking a model | The CLI’s own model, or any OpenRouter model per agent with one key (OpenRouter), or models on your own servers. |
| Several accounts of one CLI | CLI accounts: separate Claude Code and Codex sign-ins, plan meters, and Continue on another account when a limit is near. |
| Scheduled or recurring prompts | Schedule in a card’s menu sends a prompt every day at a set time; the queue lines up the next prompts for a busy agent. See Prompt queue & daily prompts. |
| A lead agent that hands out work | Arrows between cards: a lead can message, wait for and change other agents. See Squads. |
| Approvals from the phone | Remote access: the whole app in your phone’s browser — statuses, terminals, prompts, a chime when an agent needs you. Nothing to install. |
| Voice input | Dictation that runs entirely on your computer, with a global hotkey. |
| Cost tracking | Usage & costs: tokens and cost per agent, per model and per account; a Budget card pauses a workspace’s agents at a token limit. |
| Project instructions | One instructions file: keep AGENTS.md, NeuroSquad mirrors it to CLAUDE.md, GEMINI.md and QWEN.md. |
Rebuild your setup
Install NeuroSquad and create a free account. The setup wizard checks for an agent CLI and Git and can install them for you.
NeuroSquad never writes your CLI configs. Your sign-ins, settings, MCP servers and skills keep working; each card adds its own layer on top. Conversations stay in each CLI’s own history, so a session started in a card can be continued from a plain terminal and back.
Press + next to Workspaces and pick the project folder. For a remote machine, choose WSL or an SSH host in the same dialog.
Open the add menu (the arrow next to Add Claude Code). Pick a CLI for a quick start, or New agent… (provider, model, role) to set its model and role up front. Turn on Isolate in git worktree first if agents should not share files.
Drag from one card to another. An arrow from a lead to a helper lets the lead give it tasks; an arrow to a terminal, browser or note gives the agent tools for it. See Arrows give abilities.
In the canvas tools menu, Save as template… keeps the cards, layout, arrows and frames. Next time, New workspace from template… gives you the same setup in another folder.
Start from a ready-made template
You don’t have to start from an empty canvas. The template marketplace has ready squads; Use template imports one. Templates also import from a file or a link, either as a new workspace or added to the one that is open — see Workspace templates.
A template brings cards, arrows and setup, never conversations, keys or sign-ins. Agents in it that ask for their own worktree get one only if the folder is a git repository.
What to expect on day one
- Statuses come from the CLIs themselves. For most agents NeuroSquad reads the CLI’s own hooks, so “needs your input” shows what the agent is asking — the command or the question.
- Notifications are one per agent. A finished turn chimes once; a question stays until you answer. Your phone gets the same chime when it is connected.
- Nothing runs in the background without the app. Agents live while NeuroSquad is open; close it and their sessions resume next time.
Prefer the terminal? NeuroSquad also has an open-source terminal version, nsq, on GitHub .