DOCUMENTATION

SquadCoder docs

Everything you need to install, configure and master SquadCoder — the open-source AI coding agent that runs like a real engineering team.

01 Introduction

Most AI coding tools are one agent talking to one model. When a task gets complex, you babysit it prompt-by-prompt. SquadCoder runs like a real engineering org: an orchestrator breaks your request into tasks and spins up the right specialists — in parallel — then a CTO quality gate loops fixes until every check is green.

It reads and writes code, runs commands, manages Git, searches your codebase semantically, and keeps a persistent memory of your whole project. Everything ships pre-configured and works on first launch — with zero configuration. SquadCoder is MIT-licensed and built on opencode.

Open coreMIT-licensed. Free to use, modify and distribute. No telemetry you didn’t ask for.
Bring your own modelAnthropic, OpenRouter, OpenAI, Gemini, or any OpenAI-compatible endpoint. You pay your provider directly.
Local-firstYour code and context stay on your machine. The only network calls are to the model API you configure.

02 Installation

Desktop app (recommended)

Download from GitHub Releases and run the installer.

PLATFORMFILESIZE
Windows x64SquadCoder-desktop-win-x64-installer.exe~214 MB
Windows x64 (portable)SquadCoder-desktop-win-x64-portable.zip~165 MB
macOS / Linux desktopComing soon

CLI

Prefer the terminal? Download the standalone CLI, extract it, add it to your PATH, and run squadcoder from any project directory.

PLATFORMFILESIZE
Windows x64SquadCoder-cli-windows-x64.zip~48 MB
macOS / LinuxBuild from source (below)

VS Code extension

SquadCoder ships a VS Code extension for Windows, macOS and Linux — available from the repository.

From source

Requires Bun ≥ 1.3.11 — works on Windows, macOS and Linux.

$ git clone https://github.com/squadcodercom/squadcoder.git
$ cd squadcoder
$ bun install
$ bun run dev # desktop app + engine (hot reload)
$ bun run cli # CLI only

03 Quick start

Launch SquadCoder, add a model provider with your API key, open a project folder, and start chatting.

1. Launch SquadCoder (desktop app or squadcoder CLI)
2. Add a model — bring your own API key (Anthropic, OpenRouter, or any OpenAI-compatible provider)
3. Open a project folder and start chatting
4. Switch to Team Mode for anything multi-step

Tip: You don’t need to configure agents, skills, or MCPs. Everything ships pre-configured and works out of the box. Just add your API key and go.

04 Team Mode

For anything multi-step, switch to Team Mode. An orchestrator breaks your request into tasks and spins up the right specialists — in parallel — then a CTO quality gate loops fixes until every check is green. You get one coherent, verified result, not a pile of raw agent output.

You: “Add user authentication with OAuth”
               |
       [ Orchestrator ]
     /   |    |    \
 Architect  Dev1  Dev2  Security
     \   |    |    /
    [ CTO Quality Gate ]
     loops until green
           |
    [ Verified Result ]

The team ships pre-configured with sensible roles and model assignments. You can customize roles, add your own agents, or change which model each agent uses — but you don’t have to.

05 Models & providers

SquadCoder is model-agnostic — bring your own API key from any major provider. Add a provider once and every agent, skill and Team Mode role can use it.

AnthropicClaude models via your Anthropic API key.
OpenRouterOne key, hundreds of models routed through OpenRouter.
OpenAIGPT models via your OpenAI API key.
GeminiGoogle Gemini models.
OpenAI-compatibleAny endpoint that speaks the OpenAI API — local models, gateways, proxies.

Add a key in Settings → Providers → Add key, or in squadcoder.json. Use {env:VAR} placeholders so tokens never live in the file.

06 Skills & MCPs

SquadCoder ships 820+ skills ready on first launch — design systems, code review, document generation, ad platforms (Google / Meta / TikTok) and more. Skills auto-fire when relevant; no manual activation.

Built-in MCP servers

GitHub, Playwright (browser testing), web search and docs lookup are built in — plus any MCP server you add, local or remote.

Add your own

Drop a SKILL.md into .squadcoder/skills/ to teach your agents a new capability, or register an MCP server in squadcoder.json.

07 Memory & indexing

Persistent memory means SquadCoder never forgets your project — rules, architecture decisions and past fixes survive across sessions and improve over time.

A semantic codebase index lets you find code by meaning, not just text. It uses a bundled embedding model that runs fully offline — no cloud, no API cost. Ask “where do we handle auth?” and get the right files.

08 Configuration

SquadCoder works with zero config, but everything is customizable via squadcoder.json:

Model providersAnthropic, OpenRouter, OpenAI, Gemini, or any OpenAI-compatible endpoint.
MCP serversAdd any MCP server (local or remote). GitHub, Playwright, web search and docs lookup are built in.
SkillsDrop a SKILL.md into .squadcoder/skills/ to teach your agents new capabilities.
Agent rolesCustomize the Team Mode roster, or create your own specialist agents.
SecretsUse {env:VAR} or ${VAR} placeholders. Never hard-code tokens.

09 CLI reference

Run squadcoder from any project directory to open the agent in your terminal.

$ squadcoder # open the agent in the current project
$ squadcoder --help # list commands and flags
 
# inside a session:
/goal <objective> # run until the objective is verified (judge-gated)
/team # switch to Team Mode for multi-step work

Everything ships pre-configured — you can start with a single command and add providers, skills or MCPs later.

Ready to install?

Desktop app, CLI, or VS Code extension — all from GitHub Releases.