CLI Reference

The Gambi CLI provides commands for managing hubs, rooms, and participants.

All commands support interactive mode — run without flags in a terminal and you’ll be guided through each option step by step. Flags still work for scripting and automation.

If you’re coming from the old package and binary names, read Migrate from Gambiarra.

Installation

curl -fsSL https://raw.githubusercontent.com/arthurbm/gambi/main/scripts/install.sh | bash

irm https://raw.githubusercontent.com/arthurbm/gambi/main/scripts/install.ps1 | iex

npm install -g gambi
# or
bun add -g gambi

The published gambi package is a wrapper that installs only the matching platform binary for your machine.

Uninstallation

curl -fsSL https://raw.githubusercontent.com/arthurbm/gambi/main/scripts/uninstall.sh | bash

irm https://raw.githubusercontent.com/arthurbm/gambi/main/scripts/uninstall.ps1 | iex

npm uninstall -g gambi
# or
bun remove -g gambi

Interactive Mode

When you run any command without its required flags in a terminal (TTY), the CLI enters interactive mode and prompts you for each option:

┌  gambi join
│
◇  Room code:
│  ABC123
│
◆  LLM Provider:
│  ● Ollama (localhost:11434)
│  ○ LM Studio (localhost:1234)
│  ○ vLLM (localhost:8000)
│  ○ Custom URL
│
◇  Select model:
│  llama3.2
│
└  Joined room ABC123!

Interactive mode is disabled when piping input (echo "x" | gambi create), so scripts work as before.

Commands

`serve`

Start a hub server.

# Interactive — prompts for port, host, mDNS:
gambi serve

# With flags:
gambi serve [options]

Options:

Option	Description	Default
`--port`, `-p`	Port to listen on	`3000`
`--host`, `-h`	Host to bind to	`0.0.0.0`
`--mdns`, `-m`	Enable mDNS auto-discovery	`false`
`--quiet`, `-q`	Suppress logo output	`false`

Example:

gambi serve --port 3000 --mdns

`create`

Create a new room on a hub.

# Interactive — prompts for name and password:
gambi create

# With flags:
gambi create --name "Room Name" [options]

Options:

Option	Description	Default
`--name`, `-n`	Room name	Required (prompted in interactive mode)
`--password`, `-p`	Password to protect the room	None
`--hub`, `-H`	Hub URL	`http://localhost:3000`

Examples:

# Create a room interactively
gambi create

# Create with flags
gambi create --name "My Room"

# Create on a custom hub
gambi create --name "My Room" --hub http://192.168.1.10:3000

# Create a password-protected room
gambi create --name "My Room" --password secret123

`join`

Join a room and expose your LLM endpoint.

# Interactive — select provider, model, set nickname:
gambi join

# With flags:
gambi join --code <room-code> --model <model> [options]

Options:

Option	Description	Default
`--code`, `-c`	Room code to join	Required (prompted in interactive mode)
`--model`, `-m`	Model to expose	Required (prompted in interactive mode)
`--endpoint`, `-e`	Local LLM endpoint URL used for probing and inference	`http://localhost:11434`
`--network-endpoint`	Network-reachable URL to publish to the hub	Auto-detected when needed
`--nickname`, `-n`	Display name	Auto-generated
`--header`	Auth header in the format `Header=Value`	None
`--header-env`	Auth header in the format `Header=ENV_VAR`	None
`--password`, `-p`	Room password (if protected)	None
`--hub`, `-H`	Hub URL	`http://localhost:3000`
`--no-specs`	Don’t share machine specs	`false`
`--no-network-rewrite`	Disable automatic localhost-to-LAN rewrite for remote hubs	`false`

The CLI automatically probes your local endpoint to detect available models and protocol capabilities (Responses API vs Chat Completions).

When the hub is remote and your local endpoint is loopback-only (for example http://localhost:11434), Gambi tries to publish a LAN-reachable URL automatically. In interactive mode, the CLI explains the rewrite and lets you confirm or override it. Use --network-endpoint when you want to publish a specific URL yourself.

In interactive mode, you’ll select your LLM provider from a list (Ollama, LM Studio, vLLM, or custom URL), optionally add auth headers, and then choose from the detected models.

Examples:

# Join interactively — guided through all options
gambi join

# Join with Ollama
gambi join --code ABC123 --model llama3

# Join with LM Studio
gambi join --code ABC123 \
  --model mistral \
  --endpoint http://localhost:1234

# Join a remote hub and publish an explicit LAN URL
gambi join --code ABC123 \
  --hub http://192.168.1.10:3000 \
  --model llama3 \
  --endpoint http://localhost:11434 \
  --network-endpoint http://192.168.1.25:11434

# Join with custom nickname
gambi join --code ABC123 \
  --model llama3 \
  --nickname "alice-4090"

# Join a remote provider securely
export OPENROUTER_AUTH="Bearer sk-or-..."
gambi join --code ABC123 \
  --model meta-llama/llama-3.1-8b-instruct:free \
  --endpoint https://openrouter.ai/api \
  --header-env Authorization=OPENROUTER_AUTH

# Join a password-protected room
gambi join --code ABC123 \
  --model llama3 \
  --password secret123

`list`

List available rooms on a hub.

# Interactive — prompts for hub URL and output format:
gambi list

# With flags:
gambi list [options]

Options:

Option	Description	Default
`--hub`, `-H`	Hub URL	`http://localhost:3000`
`--json`, `-j`	Output as JSON	`false`

Example:

gambi list
# Output:
# Available rooms:
#   ABC123  My Room
#     Participants: 3
#   XYZ789  Test Room
#     Participants: 1

gambi list --json

`monitor`

Open the TUI to monitor rooms in real-time.

gambi monitor [options]

Options:

Option	Description	Default
`--hub`, `-H`	Hub URL	`http://localhost:3000`

The monitor shows participants, their status (online/offline), and a live activity log of events (joins, requests, errors) via SSE.

Tip: The standalone CLI currently shows help when you run gambi without arguments. Use gambi monitor to open the TUI.

Supported Providers

Gambi works with any endpoint that exposes OpenResponses or OpenAI-compatible chat/completions:

Provider	Default Endpoint	Protocols
Ollama	`http://localhost:11434`	Responses API, Chat Completions
LM Studio	`http://localhost:1234`	Responses API, Chat Completions
LocalAI	`http://localhost:8080`	Responses API, Chat Completions
vLLM	`http://localhost:8000`	Responses API, Chat Completions

For cloud providers (OpenRouter, Together AI, Groq, etc.), see the Remote Providers guide.