Agents Squads CLI
The squads CLI runs teams of agents defined in markdown files in git: a lead briefs, workers act, a verifier checks.
- 56 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- Agents run without constant supervision, write files and commit to the repository themselves.
- Every run spends tokens and keys of your providers.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/agents-squadsDetects 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 without constant supervision, write files and commit to the repository themselves.
- Every run spends tokens and keys of your providers.
This is third-party code. Review the repository files before installing.
What it does
Agents Squads describes a team of AI agents with plain markdown files in a git repository. A lead briefs, workers act, a verifier checks the result, and feedback from each cycle feeds the next. There is no database, server or custom markup language: git is state and audit trail. `squads init` creates an `.agents/` directory with starter squads, `squads run` starts an agent or a squad conversation, `squads inbox` lists everything waiting for your decision. Agents run through the native model CLIs: `claude`, `gemini`, `aider`, `opencode`. Mixed teams therefore work at once: a cheap model scans, a strong one builds, a mid-tier one verifies.
Who it is for. Teams and solo developers on Claude Code who want repeatable agent processes with memory and a log in git.
Good fit when
- A research squad collects topic material every morning, and you decide from `squads inbox`.
- One task needs different models: a cheap one for tagging, a strong one for building.
- Agent work must be visible in git history with clear commits.
- Squad settings live in the repository beside the project, and a newcomer gets them with `git clone`.
Not a fit when
- A single one-off prompt to an agent is enough: squad machinery is overkill.
- Automatic agent commits into the repository are unacceptable.
- Claude Code is not installed: it is the default provider.
Example request
Create a research squad, run the analyst and show in the inbox what waits for my decision.Limitations
You need Node.js 20 or newer, Git and Claude Code signed in with `claude /login`; other providers are optional. Agents run in your environment with your keys, and the demo run auto-commits its result locally. `squads doctor` checks the environment and says what is missing. Anonymous telemetry covers only command names, versions and error classes, and is disabled with `DO_NOT_TRACK=1` or `SQUADS_TELEMETRY_DISABLED=1`. Cloud login and `--cloud` runs stay hidden until an auth address is configured.
How to disable. Uninstall the package with `npm uninstall -g squads-cli` and remove the `.agents/` directory and agent commits from the repository by hand.
FAQ
How do I install squads?
Run `npm install -g squads-cli`, then `git init` and `squads init` in a new folder. Check the environment with `squads doctor`.
How is a squad different from one agent?
A squad assembles a lead, workers and a verifier with shared memory and goals. Feedback from the previous cycle is mixed into the next run.
Which models can be connected?
Those whose CLIs are installed: `claude`, `gemini`, `aider`, `opencode` and more. Routing can be set for each agent separately.
What is stored in `.agents/`?
A business brief, the immutable `config/SYSTEM.md` rules, squad descriptions with one markdown file per agent and a `memory/` directory with goals, learnings and feedback.
What does `squads run demo hello-world` show?
The agent prints a greeting, writes a state file and makes a local commit. If all went well, the install works.
Discussion
No questions or reviews yet.
Related
CLI toolsmulti-agents runs a team of coding agents on a task and measures whether it beat a single agent and what it cost.
Squad creates coding-agent teams with roles and shared repository memory for distributing tasks and retaining decisions.
Runs sub-agents from Markdown files in the .agents folder on backends such as Codex, Claude Code, Cursor CLI and Gemini.
A CLI to run several coding agents: one config, a session console, local memory and a set of engineering skills
dotagents is a Sentry CLI that wires skills, MCP servers, hooks and plugins from one agents.toml into each agent's folders.
A scaffold of personas, rules and skills in plain Markdown: Maestro hands work to different models and spares your main plan
README in short
Squads and memory live in `.agents/` beside the code, and before every run an agent loads a context cascade tuned to its role. The README lists docs on philosophy, architecture, configuration, running, commands and providers.