Docs — Open CLI

Quick Start

Install Open CLI globally and run your first command in under 60 seconds.

# 1. Install
$ npm install -g github:rythmmcosta/open-cli

# 2. Connect AI provider
$ opencli auth

# 3. Run a prompt
$ opencli "list all .ts files changed today"

# 4. Interactive REPL
$ opencli

Authentication

Open CLI supports four AI providers. Run opencli auth to connect any of them.

Supported Providers

Anthropic — compatible text models; copy the current model ID from your provider account
OpenAI — GPT-4o, gpt-4o-mini, o1-preview, o1-mini
Google Gemini — gemini-2.0-flash, gemini-2.0-flash-lite, gemini-1.5-pro
Ollama — Any local model (llama3.2, codellama, mistral, etc.)

# Interactive setup
$ opencli auth

# Or use environment variables
$ export ANTHROPIC_API_KEY="sk-ant-..."
$ export OPENAI_API_KEY="sk-..."
$ export GEMINI_API_KEY="AIza..."

CLI Flags

All available flags for the opencli command:

Flag	Short	Description
`--model <name>`	`-m`	AI model to use
`--skill <name>`	`-s`	Built-in skill to activate
`--yes`	`-y`	Auto-approve low-risk commands
`--dry-run`	—	Preview tool calls without executing
`--profile <name>`	`-p`	Apply a saved config profile
`--show-cost`	—	Display token usage and estimated cost
`--no-color`	—	Disable terminal colors
`--verbose`	`-v`	Show raw tool call details
`--context <file>`	`-c`	Inject a file into the prompt context
`--browser`	—	Enable browser automation tools
`--notify`	—	Send notification when task completes
`--version`	`-V`	Show version
`--help`	`-h`	Show help text

# Examples
$ opencli --model gpt-4o "explain this code"
$ opencli --skill python --yes "add tests"
$ opencli --dry-run "delete all node_modules"
$ opencli -c package.json "update to latest deps"

Subcommands

opencli auth

Interactive menu to add/remove/test AI provider API keys.

opencli skills

List all 35 built-in skills with descriptions.

opencli mcp

List all 8 built-in MCP servers and their tools.

opencli config

Show current configuration (model, skill, providers, etc.).

opencli browser <action>

$ opencli browser install # Install Playwright
$ opencli browser screenshot <url> # Take screenshot
$ opencli browser verify <url> # Verify page

opencli agent <subcommand>

$ opencli agent list # List agents
$ opencli agent types # Show types
$ opencli agent run <name> # Run now
$ opencli agent background <name> # Background run
$ opencli agent delete <name> # Delete

opencli notify <subcommand>

$ opencli notify setup # Configure
$ opencli notify test # Test send
$ opencli notify send "Done" --level success

REPL Slash Commands

When running in interactive mode (opencli), these slash commands are available:

Command	Description
`/auth`	Open auth wizard
`/model <name>`	Switch AI model
`/skill <name>`	Switch skill
`/skills`	List all skills
`/profile <name>`	Apply a saved profile
`/mcp`	List MCP servers
`/browser <action>`	Browser commands
`/agent <subcommand>`	Agent management
`/notify <subcommand>`	Notification commands
`/dry-run`	Toggle dry-run mode
`/auto`	Toggle auto-approve
`/clear`	Clear the screen
`/reset`	Reset conversation
`/history`	Show message history
`/cost`	Show token usage
`/help`	Show all commands
`/exit`	Exit REPL

Configuration

Config is stored at ~/.config/opencli/config.json.

{
  "defaultModel": "YOUR_MODEL_ID",
  "activeSkill": "default",
  "autoApprove": false,
  "dryRun": false,
  "contextWindow": 20,
  "showUsage": false,
  "providers": {
    "anthropic": { "apiKey": "sk-ant-..." },
    "ollama": { "baseUrl": "http://localhost:11434" }
  }
}

Profiles

Profiles let you save and switch between sets of config: model + skill + system prompt.

"profiles": {
  "frontend": {
    "model": "YOUR_MODEL_ID",
    "skill": "vuejs"
  },
  "backend": {
    "model": "gpt-4o",
    "skill": "nodejs",
    "system": "Senior backend engineer. Be concise."
  }
}

$ opencli --profile frontend "add dark mode toggle"
❯ /profile backend

Environment Variables

API keys can be set via environment variables instead of the config file:

export ANTHROPIC_API_KEY="sk-ant-..."
export OPENAI_API_KEY="sk-..."
export GEMINI_API_KEY="AIza..."

# Add to ~/.bashrc or ~/.zshrc to persist

MCP Servers

All 18 MCP servers are built-in. Use opencli mcp to list them.

filesystem — read_file, write_file, list_files, search_files
shell — execute_bash
git — git_command
browser — navigate, screenshot, click, fill_form, verify, extract
http — http_request (GET/POST/PUT/DELETE)
database — db_query, db_schema
docker — docker_list, docker_logs, docker_exec, docker_compose
search — web_search, fetch_page

Security

Open CLI uses a 3-tier risk scoring system to keep your system safe.

Risk Levels

LOW — Read-only commands. Auto-executed with -y.
MEDIUM — Elevated access (sudo, global installs). Always prompts [Y/n].
HIGH — Destructive (rm -rf, DROP TABLE). Must type "yes" in full.

Best Practices

Use --dry-run first on cleanup/delete operations
Only use -y in trusted, known project directories
API keys are stored with 0600 permissions — never share your config
Reading .env, *.pem, or ~/.ssh/* auto-upgrades to MEDIUM risk

Termux (Android)

Full Open CLI works on Termux, except browser automation (Playwright requires a desktop OS).

# Step 1: Install deps
$ pkg update && pkg install nodejs git python make clang

# Step 2: Clone and install
$ git clone https://github.com/rythmmcosta/open-cli.git
$ cd open-cli && npm install && npm run build && npm link

# Step 3: Configure
$ opencli auth

See the full Termux installation guide →

Documentation