#Quickstart

This guide gets you from zero to a running Wingman setup with a gateway, configured provider, and bundled agents.

#0) Install Wingman

Choose one option:

curl -fsSL https://getwingmanai.com/install.sh | bash

or

npm install -g @wingman-ai/gateway

The install script runs wingman init for you. If you use npm, continue to step 1.

#1) Initialize a workspace

From the root of your project:

wingman init

This creates .wingman/wingman.config.json and installs bundled agents in .wingman/agents/.

You can choose bundled agents during setup:

wingman init --agents main,coding,stock-trader

#2) Configure a model provider

Either set an env var or store credentials:

# Example: OpenAI
wingman provider login openai

# Example: OpenAI Codex subscription auth
wingman provider login codex

Wingman stores credentials in ~/.wingman/credentials.json by default. You can also configure via env vars.

#3) Start the gateway

wingman gateway start

Defaults:

• Gateway WebSocket: ws://127.0.0.1:18789/ws
• Control UI: http://127.0.0.1:18790/

#4) Invoke an agent

wingman agent --agent main "Summarize the repo architecture."

Use --local to bypass the gateway and run locally:

wingman agent --agent main --local "Refactor the config loader."

#Next steps

• Configure providers and model strings in detail.
• Create custom agents and sub-agents.
• Enable adapters like Discord or Voice.
• Add MCP servers and skills.