Skip to main content
AI coding assistants can write code but can’t click buttons or test UIs. AskUI fixes that.

​
MCP Integration

from fastmcp import Client
from fastmcp.mcp_config import MCPConfig
from askui.agent import VisionAgent
from askui.models.shared.tools import ToolCollection
from askui.tools.mcp.config import StdioMCPServer

mcp_config = MCPConfig(
    mcpServers={
        "askui": StdioMCPServer(
            command="python",
            args=["-m", "askui.tools.mcp.servers.stdio"]
        ),
    }
)

mcp_client = Client(mcp_config)
tools = ToolCollection(mcp_client=mcp_client)

with VisionAgent() as agent:
    agent.act("Test the login flow", tools=tools)

​
Direct Integration

with VisionAgent() as agent:
    agent.act("""
        Open localhost:3000
        Click "Sign Up"
        Fill registration form
        Verify success message
    """)

​
Direct Tool Access

with VisionAgent() as agent:
    # Mouse
    agent.tools.os.click("left", 2)  # Double-click
    agent.tools.os.mouse_move(100, 200)

    # Keyboard
    agent.tools.os.keyboard_tap("v", modifier_keys=["control"])
    agent.tools.os.keyboard_type("Hello")

    # Clipboard
    agent.tools.clipboard.copy("text")
    text = agent.tools.clipboard.paste()

    # Browser
    agent.tools.webbrowser.open_new("https://example.com")

​
Multi-Display

with VisionAgent(display=2) as agent:
    agent.act("Open dashboard on second monitor")