Skip to content

interface-cli

Directory actions

More options

Directory actions

More options

Latest commit

 

History

History

interface-cli

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
src
src
 
 
README.md
README.md
 
 
package.json
package.json
 
 
project.json
project.json
 
 
tsconfig.app.json
tsconfig.app.json
 
 
tsconfig.json
tsconfig.json
 
 

README.md

Browcall CLI (@ducy23061999/browcall-cli)

npm version license

Browcall CLI is a self-contained launcher and management tool for Browcall backend services:

  • gpt-auto-api: OpenAI/Anthropic compatible API & SSE server connecting to live browser automation.
  • mcp-gateway: Remote Model Context Protocol (MCP) HTTP/WebSocket proxy and tool router.
  • local-mcp-bridge: Local MCP stdio bridge linking local tool servers (like @modelcontextprotocol/server-filesystem) to the Gateway.

⚡ Quick Start

Option 1: Instant Execution via npx (No installation needed)

Launch all 3 backend components instantly:

npx @ducy23061999/browcall-cli@latest start --all

Launch the interactive configuration wizard:

npx @ducy23061999/browcall-cli@latest
# or
npx @ducy23061999/browcall-cli@latest interactive

Option 2: Global Installation via npm / yarn

npm install -g @ducy23061999/browcall-cli
# or
yarn global add @ducy23061999/browcall-cli

Once installed globally, you can use the short command browcall:

# Start all components
browcall start --all

# Launch interactive menu
browcall

# Check service build & process status
browcall status

# Kill previous running processes & free bound ports
browcall kill

🛠 Usage & Command Reference

USAGE:
  browcall [command] [options]
  interface-cli [command] [options]
  npx @ducy23061999/browcall-cli [command] [options]

Commands

Command Description
start [services...] Start specified backend components or all if none given (gpt-auto-api, mcp-gateway, local-mcp-bridge).
interactive Launch interactive wizard to configure project directories and component ports.
generate-config Build mcp-config.json from template for a target project directory.
status View running PID and build status of all backend components.
kill / stop Automatically terminate previous running Browcall instances and release bound ports.
help Display CLI help menu.

Command Options & Flags

Flag Parameter Default Description
--all N/A N/A Launch all 3 backend services simultaneously.
--api N/A N/A Launch gpt-auto-api service only.
--gateway N/A N/A Launch mcp-gateway service only.
--bridge N/A N/A Launch local-mcp-bridge service only.
--project-dir <path> process.cwd() Target execution directory for filesystem MCP tools.
--port-api <port> 8766 Set HTTP API port for gpt-auto-api.
--port-api-ws <port> 8765 Set Browser Extension WebSocket port for gpt-auto-api.
--port-gateway <port> 8767 Set HTTP endpoint port for mcp-gateway.
--port-ws <port> 8768 Set WebSocket router port for mcp-gateway & bridge target.
--ngrok N/A N/A Expose MCP Gateway via public ngrok HTTPS tunnel for online MCP importing.
--ngrok-authtoken <token> N/A Optional ngrok authtoken.
--generate-config N/A N/A Auto-generate mcp-config.json inside --project-dir.
--env-file <path> N/A Path to custom .env file to load environment variables from.

💡 Practical Examples

1. Start all components for a specific target project

browcall start --all --project-dir /Users/username/Projects/my-app --generate-config

2. Run with custom ports

browcall start --all \
  --port-api 9000 \
  --port-api-ws 9001 \
  --port-gateway 9002 \
  --port-ws 9003

3. Generate mcp-config.json template only

browcall generate-config --project-dir /Users/username/Projects/my-app

🤖 Connecting AI Clients & Claude Code CLI

1. Claude Code CLI (Anthropic API Compatible)

Set environment variables to route Claude Code completions through gpt-auto-api:

export ANTHROPIC_BASE_URL="http://localhost:8766"
export ANTHROPIC_API_KEY=""

# Run Claude Code CLI
claude

2. OpenAI Compatible API Clients

  • Base URL: http://localhost:8766/v1
  • Endpoint: POST /v1/chat/completions

3. MCP Clients (Claude Desktop, Cursor, AI Agents)

  • MCP HTTP Endpoint: http://localhost:8767/mcp
  • Live SSE Endpoint: http://localhost:8767/sse
  • Gateway WebSocket: ws://localhost:8768
  • Public Ngrok Endpoints (--ngrok): https://xxxx.ngrok-free.app/mcp & https://xxxx.ngrok-free.app/sse

For step-by-step setup guides, see the Claude Code & AI Agent Guide.

🧹 Automatic Cleanup & Port Release

Browcall CLI automatically handles process lifecycle and signal management:

  • When starting services, any previous zombie instances or stale port bindings on 8766, 8765, 8767, or 8768 are safely terminated.
  • When pressing Ctrl+C or exiting the CLI, all child processes are killed and ports are freed cleanly.
  • You can manually force cleanup anytime via browcall kill.

📄 License

MIT © Tran Duc Y