Brigade is a polished terminal interface for working with the world's leading language models — Anthropic Claude, OpenAI GPT, Google Gemini, Groq, xAI, DeepSeek, Mistral, OpenRouter, Cerebras — and your own local models via Ollama or any OpenAI-compatible endpoint.

It runs in two modes: a single-process chat TUI (the default) or as a headless WebSocket gateway you can connect multiple thin clients to.

Install

npm install -g @spinabot/brigade

Requires Node.js 20 or newer.

Quick start

brigade

First launch walks you through three steps:

1. Pick a provider.
2. Connect it (paste an API key, or scan your local Ollama).
3. Choose a default model.

That's it. Subsequent launches resume right where you left off.

Commands

Brigade is a single binary with subcommands. brigade on its own is shorthand for brigade chat.

| Command | What it does | |--------------------|---------------------------------------------------------------| | brigade | Start the chat TUI (default) | | brigade chat | Start the chat TUI | | brigade gateway | Run the headless WebSocket gateway, no TUI | | brigade connect | Open a TUI that talks to a running gateway | | brigade onboard | Re-run the provider/model setup wizard | | brigade doctor | Health-check Node, config, providers, prompts, and log sink | | brigade config | Read & write ~/.brigade/config.json | | brigade --version| Print the version | | brigade --help | Print the full help text |

brigade gateway

Runs Brigade as a WebSocket server with no terminal UI of its own. Useful when you want a long-lived agent process you can attach and detach clients to.

brigade gateway --port 7777 --host 127.0.0.1 --verbose

| Flag | Default | Notes | |------------------|---------------|----------------------------------------------| | --port N | 7777 | Listen port (also BRIGADE_PORT env var) | | --host A | 127.0.0.1 | Bind address | | --verbose | off | Stream a one-line summary of every event | | --quiet | off | Suppress the live console stream | | --log-level X | info | debug, info, warn, error |

brigade connect

Attaches a TUI to a running gateway. Same chat experience as brigade, but the agent runs in the gateway process — so you can disconnect, walk away, reconnect later, and pick up where you left off.

brigade connect --host 127.0.0.1 --port 7777

| Flag | Default | Notes | |----------------|---------------|--------------------------------------| | --host A | 127.0.0.1 | Gateway host to connect to | | --port N | 7777 | Gateway port | | --timeout MS | 60000 | Per-request timeout |

brigade doctor

Runs a health check across Node version, your ~/.brigade/ directory, config, configured providers, log sink, prompt files, and (optionally) a running gateway. Exits 0 if everything passes, 1 if anything fails.

brigade doctor
brigade doctor --gateway ws://127.0.0.1:7777

brigade config

Read and write the local config without opening the TUI.

brigade config list
brigade config get defaultProvider
brigade config set defaultProvider anthropic
brigade config unset thinkingLevel

Known keys: defaultProvider, defaultModelId, thinkingLevel, fallbackProvider, fallbackModelId, installedAt.

In-chat commands

When you're in the chat TUI (whether via brigade or brigade connect):

| Command | What it does | |---------------------|-------------------------------------------------------------| | /model | Switch to a different configured model (picker) | | /model <id> | Switch directly by model id | | /provider | Add a new provider mid-session — no restart required | | /thinking <level> | Adjust reasoning effort (off, minimal, low, medium, high) | | /compact | Force a context compaction now | | /help | Show all commands | | /exit | Quit Brigade |

Keyboard:

• Enter — send (mid-turn submits steer the model without aborting)
• Ctrl+C — stop the current response (doesn't exit)
• Ctrl+D — quit
• ↑ / ↓ — message history

Supported providers

Out of the box: Anthropic, OpenAI, Google Gemini, OpenRouter, Groq, xAI, Cerebras, DeepSeek, Mistral, Ollama (local), and Custom OpenAI-compatible endpoints (Together AI, Fireworks, vLLM, on-prem gateways, anything that speaks /v1/chat/completions).

You can mix and match — connect three providers, switch between their models with /model, and Brigade keeps your conversation context across the switch.

Privacy

Brigade is a local CLI. Your API keys never leave your computer; they're stored in your home directory and used only to talk to the providers you connect. No telemetry, no analytics, no cloud component.

For Ollama and Custom endpoints, requests stay entirely on your network.