Documentation — PKA CODE

Web Chat — Server Setup

Launch the PKA CODE web server and chat with AI instantly. One command does everything.

1. Launch Web Chat

Install the CLI globally, then run:
pkacodeweb
The server starts on http://localhost:3721 and your browser opens automatically.

2. Chat Instantly

The chat opens at:
http://localhost:3721/chat
Start chatting with DeepSeek V4 Flash immediately — no API key needed.

3. Instant Mode (No Key)

Chat instantly with up to 20 messages per hour. The server uses its own API key for instant access.

4. Unlimited Mode (Your Key)

For unlimited usage, add your own OpenRouter API key in Settings. One free key gives access to all providers.

Manual Start (Alternative)

You can also start the server manually:
cd server && node index.js
Then open http://localhost:3721/chat in your browser.

Phone Access

The server broadcasts on your local network. Access from your phone at:
http://<your-ip>:3721/chat
Both devices must be on the same WiFi.

View All Features

Installation

Install globally via npm. Node.js 18+ required.

Install globally npm

npm install -g pkacode-cli

Start interactive mode terminal

pkacode

Single prompt mode terminal

pkacode -p "explain this function"
pkacode "build a todo app"
pkacode -p "read package.json and explain it"

Node ≥ 18.0.0 Windows Linux macOS

System requirements

Minimal dependencies. Runs on any modern terminal.

Platform	Requirement	Notes
All	Node.js 18+	LTS recommended
Windows 10+	Windows Terminal (recommended)	Clipboard built-in via `clip`
Linux	Any modern distro	Install `xclip` or `wl-clipboard`
macOS 12+	Intel & Apple Silicon	Clipboard built-in via `pbcopy`

Authentication

Set your FREE API key from any supported provider.

pkacode auth login gemini
pkacode auth login groq
pkacode auth login openrouter

Switch provider terminal

pkacode provider set gemini
pkacode provider set groq
pkacode provider set openrouter

CLI commands

All available commands for the pkacode CLI.

Command	Description
`pkacode`	Start interactive chat mode
`pkacode "prompt"`	Single-prompt mode
`pkacode -p "prompt"`	Headless print mode
`pkacode --help`	Show help
`pkacode --version`	Show version
`pkacodeweb`	Start web server and open chat in browser
`pkacode theme [dark\|light]`	Switch color theme
`pkacode auth login <provider>`	Set API key
`pkacode provider set <name>`	Switch provider
`pkacode model set <name>`	Switch model
`pkacode models list`	List all models
`pkacode config`	Show config
`pkacode config set <key> <value>`	Set config value
`pkacode permissions`	Show permission rules
`pkacode permissions reset`	Reset to defaults
`pkacode sessions list`	List saved sessions
`pkacode resume [session]`	Resume a session
`pkacode sandbox`	Toggle sandbox mode

In-chat commands

Slash commands available during interactive chat.

Command	Alias	Description
`/theme`	—	Switch color theme
`/session`	`/info`	Show session ID, model, turns, tokens
`/new`	—	Start a fresh session
`/resume`	—	Interactive picker to continue a session
`/sessions`	—	List all saved sessions
`/permissions`	`/perm`	View and change permission mode
`/exit`	`/quit`	Exit the assistant
`/clear`	—	Clear conversation history
`/help`	—	Show help
`/commands`	—	Show all PKA CLI commands
`/model`	—	Show or switch active model
`/models`	—	List available models
`/provider`	—	Show or switch AI provider
`/auth`	—	Login/logout from providers
`/tokens`	—	Show token usage
`/config`	—	Show current configuration
`/reset`	—	Reset configuration to defaults
`/sandbox`	—	Toggle sandbox preference
`/multiline`	—	Enter multi-line input mode

Configuration

Config is stored as JSON. Accessible and editable.

Config paths

Windows: %APPDATA%\pka\config.json
Linux / macOS: ~/.config/pka/config.json

Example config.json JSON

{
  "provider": "gemini",
  "model": "gemini-2.0-flash",
  "apiKeys": {
    "gemini": "AIzaSy...",
    "groq": "gsk_...",
    "openrouter": "sk-or-..."
  },
  "permissionMode": "default",
  "enabledTools": ["bash", "fs"],
  "sandbox": false,
  "maxTokens": 4096,
  "temperature": 0.3,
  "stream": true
}

Config keys

provider — Active AI provider
model — Active model name
apiKeys — Stored API keys
permissionMode — default, auto-edit, plan, or yolo
enabledTools — Allowed tool categories
sandbox — Sandbox toggle
maxTokens — Max response tokens
temperature — Model temperature (0.0–1.0)
stream — Streaming responses

Development

Build from source and contribute to the project.

Clone and build bash

git clone https://github.com/Pisethz/pkacode-cli.git
cd pkacode-cli
npm install
npm run build
npm link
pkacode

Dev mode with hot reload bash

npm run dev

TypeScript Commander.js Chalk Execa Ora MIT License