wcgw 1.5.0__py3-none-any.whl → 1.5.2__py3-none-any.whl

wcgw/client/anthropic_client.py

@@ -223,9 +223,10 @@ def loop(
             input_schema=GetScreenInfo.model_json_schema(),
             name="GetScreenInfo",
             description="""
-- Get display information of an OS running on docker using image "ghcr.io/anthropics/anthropic-quickstarts:computer-use-demo-latest"
-- If user hasn't provided docker image id, check using `docker ps` and provide the id.
 - Important: call this first in the conversation before ScreenShot, Mouse, and Keyboard tools.
+- Get display information of a linux os running on docker using image "ghcr.io/anthropics/anthropic-quickstarts:computer-use-demo-latest"
+- If user hasn't provided docker image id, check using `docker ps` and provide the id.
+- If the docker is not running, run using `docker run -d -p 6080:6080 ghcr.io/anthropics/anthropic-quickstarts:computer-use-demo-latest`
 - Connects shell to the docker environment.
 - Note: once this is called, the shell enters the docker environment. All bash commands will run over there.
 """,
@@ -234,26 +235,22 @@ def loop(
             input_schema=ScreenShot.model_json_schema(),
             name="ScreenShot",
             description="""
-- Capture screenshot of an OS running on docker using image "ghcr.io/anthropics/anthropic-quickstarts:computer-use-demo-latest"
-- If user hasn't provided docker image id, check using `docker ps` and provide the id.
-- Capture ScreenShot of the current screen for automation.
+- Capture screenshot of the linux os on docker.
 """,
         ),
         ToolParam(
             input_schema=Mouse.model_json_schema(),
             name="Mouse",
             description="""
-- Interact with docker container running image "ghcr.io/anthropics/anthropic-quickstarts:computer-use-demo-latest"
-- If user hasn't provided docker image id, check using `docker ps` and provide the id.
-- Interact with the screen using mouse
+- Interact with the linux os on docker using mouse.
+- Uses xdotool
 """,
         ),
         ToolParam(
             input_schema=Keyboard.model_json_schema(),
             name="Keyboard",
             description="""
-- Interact with docker container running image "ghcr.io/anthropics/anthropic-quickstarts:computer-use-demo-latest"
-- If user hasn't provided docker image id, check using `docker ps` and provide the id.
+- Interact with the linux os on docker using keyboard.
 - Emulate keyboard input to the screen
 - Uses xdootool to send keyboard input, keys like Return, BackSpace, Escape, Page_Up, etc. can be used.
 - Do not use it to interact with Bash tool.

wcgw/client/computer_use.py

@@ -26,7 +26,6 @@ from ..types_ import (
 OUTPUT_DIR = "/tmp/outputs"
 TYPING_DELAY_MS = 12
 TYPING_GROUP_SIZE = 50
-TRUNCATED_MESSAGE: str = "<response clipped><NOTE>To save on context only part of this file has been shown to you.</NOTE>"
 
 Action = Literal[
     "key",
@@ -187,6 +186,7 @@ class ComputerTool:
         docker_image_id: Optional[str] = None,
         text: str | None = None,
         coordinate: tuple[int, int] | None = None,
+        do_left_click_on_move: bool | None = None,
         **kwargs: Any,
     ) -> ToolResult:
         if action == "get_screen_info":
@@ -218,7 +218,12 @@ class ComputerTool:
             )
 
             if action == "mouse_move":
-                return self.shell(f"{self.xdotool} mousemove  {x} {y}")
+                if not do_left_click_on_move:
+                    return self.shell(f"{self.xdotool} mousemove  {x} {y}")
+                else:
+                    return self.shell(
+                        f"{self.xdotool} mousemove  {x} {y} click 1",
+                    )
             elif action == "left_click_drag":
                 return self.shell(
                     f"{self.xdotool} mousedown 1 mousemove  {x} {y} mouseup 1",
@@ -402,6 +407,7 @@ def run_computer_tool(
             result = Computer(
                 action="mouse_move",
                 coordinate=(action.action.x, action.action.y),
+                do_left_click_on_move=action.action.do_left_click_on_move,
             )
         elif isinstance(action.action, LeftClickDrag):
             result = Computer(

wcgw/client/mcp_server/Readme.md

@@ -1,15 +1,10 @@
 # Claude desktop support
 
-## Setup
-
-Install xdtool
+`wcgw` enables Claude desktop app on Mac to access shell and file system in order to automate tasks, run code, etc.
 
-```sh
-brew install xdotool
+It also has a computer use feature to connect to linux running on docker. Claude can fully control it including mouse and keyboard.
 
-# On macos:
-defaults write org.x.X11 enable_test_extensions -boolean true
-```
+## Setup
 
 Update `claude_desktop_config.json` (~/Library/Application Support/Claude/claude_desktop_config.json)
 
@@ -17,8 +12,16 @@ Update `claude_desktop_config.json` (~/Library/Application Support/Claude/claude
 {
   "mcpServers": {
     "wcgw": {
-      "command": "uvx",
-      "args": ["--from", "wcgw@latest", "wcgw_mcp"]
+      "command": "uv",
+      "args": [
+        "tool",
+        "run",
+        "--from",
+        "wcgw@latest",
+        "--python",
+        "3.12",
+        "wcgw_mcp"
+      ]
     }
   }
 }
@@ -26,28 +29,17 @@ Update `claude_desktop_config.json` (~/Library/Application Support/Claude/claude
 
 Then restart claude app.
 
-### Computer use support using desktop on docker
+### [Optional] Computer use support using desktop on docker
 
-Controlling the system isn't possible yet, but you can connect to a docker container which runs a linux os with desktop.
+Computer use is enabled by default. Claude will be able to connect to any docker container with linux environment. Native system control isn't supported outside docker.
 
 First run a sample docker image with desktop and optionally VNC connection:
 
 ```sh
-docker run \
-    --entrypoint "" \
-    -p 6080:6080 \
-    -e WIDTH=1024 \
-    -e HEIGHT=768 \
-    -d \
-    ghcr.io/anthropics/anthropic-quickstarts:computer-use-demo-latest \
-    bash -c "\
-        ./start_all.sh && \
-        ./novnc_startup.sh && \
-        python http_server.py > /tmp/server_logs.txt 2>&1 & \
-        tail -f /dev/null"
+docker run -p 6080:6080 ghcr.io/anthropics/anthropic-quickstarts:computer-use-demo-latest
 ```
 
-Connect to `http://localhost:6080/vnc.html` for desktop view (VNC) of the system running in the docker.
+Connect to `http://localhost:6080/vnc.html` for desktop view (VNC) of the system running in the docker. Then ask claude desktop app to control the docker os.
 
 ## Usage
 

wcgw/client/mcp_server/server.py

@@ -10,8 +10,9 @@ from mcp.server.models import InitializationOptions
 import mcp.types as types
 from mcp.types import Tool as ToolParam
 from mcp.server import NotificationOptions, Server
-from pydantic import AnyUrl, ValidationError
+from pydantic import AnyUrl, BaseModel, ValidationError
 import mcp.server.stdio
+from .. import tools
 from ..tools import DoneFlag, get_tool_output, which_tool_name, default_enc
 from ...types_ import (
     BashCommand,
@@ -29,6 +30,7 @@ from ...types_ import (
 )
 from ..computer_use import Computer
 
+tools.TIMEOUT = 3
 
 server = Server("wcgw")
 
@@ -149,9 +151,10 @@ async def handle_list_tools() -> list[types.Tool]:
             inputSchema=GetScreenInfo.model_json_schema(),
             name="GetScreenInfo",
             description="""
-- Get display information of an OS running on docker using image "ghcr.io/anthropics/anthropic-quickstarts:computer-use-demo-latest"
-- If user hasn't provided docker image id, check using `docker ps` and provide the id.
 - Important: call this first in the conversation before ScreenShot, Mouse, and Keyboard tools.
+- Get display information of a linux os running on docker using image "ghcr.io/anthropics/anthropic-quickstarts:computer-use-demo-latest"
+- If user hasn't provided docker image id, check using `docker ps` and provide the id.
+- If the docker is not running, run using `docker run -d -p 6080:6080 ghcr.io/anthropics/anthropic-quickstarts:computer-use-demo-latest`
 - Connects shell to the docker environment.
 - Note: once this is called, the shell enters the docker environment. All bash commands will run over there.
 """,
@@ -160,29 +163,26 @@ async def handle_list_tools() -> list[types.Tool]:
             inputSchema=ScreenShot.model_json_schema(),
             name="ScreenShot",
             description="""
-- Capture screenshot of an OS running on docker using image "ghcr.io/anthropics/anthropic-quickstarts:computer-use-demo-latest"
-- If user hasn't provided docker image id, check using `docker ps` and provide the id.
-- Capture ScreenShot of the current screen for automation.
+- Capture screenshot of the linux os on docker.
 """,
         ),
         ToolParam(
             inputSchema=Mouse.model_json_schema(),
             name="Mouse",
             description="""
-- Interact with docker container running image "ghcr.io/anthropics/anthropic-quickstarts:computer-use-demo-latest"
-- If user hasn't provided docker image id, check using `docker ps` and provide the id.
-- Interact with the screen using mouse
+- Interact with the linux os on docker using mouse.
+- Uses xdotool
 """,
         ),
         ToolParam(
             inputSchema=Keyboard.model_json_schema(),
             name="Keyboard",
             description="""
-- Interact with docker container running image "ghcr.io/anthropics/anthropic-quickstarts:computer-use-demo-latest"
-- If user hasn't provided docker image id, check using `docker ps` and provide the id.
+- Interact with the linux os on docker using keyboard.
 - Emulate keyboard input to the screen
 - Uses xdootool to send keyboard input, keys like Return, BackSpace, Escape, Page_Up, etc. can be used.
 - Do not use it to interact with Bash tool.
+- Make sure you've selected a text area or an editable element before sending text.
 """,
         ),
     ]
@@ -272,7 +272,7 @@ async def main() -> None:
                             experimental_capabilities={},
                         ),
                     ),
-                    raise_exceptions=True,
+                    raise_exceptions=False,
                 )
         except BaseException as e:
             print(f"Server encountered an error: {e}", file=sys.stderr)

wcgw/client/sys_utils.py

@@ -1,6 +1,7 @@
 import subprocess
 
 MAX_RESPONSE_LEN: int = 16000
+TRUNCATED_MESSAGE: str = "<response clipped><NOTE>To save on context only part of this file has been shown to you.</NOTE>"
 
 
 def maybe_truncate(content: str, truncate_after: int | None = MAX_RESPONSE_LEN) -> str:
@@ -14,7 +15,7 @@ def maybe_truncate(content: str, truncate_after: int | None = MAX_RESPONSE_LEN)
 
 def command_run(
     cmd: str,
-    timeout: float | None = 30.0,  # seconds
+    timeout: float | None = 3.0,  # seconds
     truncate_after: int | None = MAX_RESPONSE_LEN,
     text: bool = True,
 ) -> tuple[int, str, str]:

wcgw/client/tools.py

@@ -71,7 +71,7 @@ from .openai_utils import get_input_cost, get_output_cost
 
 console = rich.console.Console(style="magenta", highlight=False, markup=False)
 
-TIMEOUT = 30
+TIMEOUT = 5
 
 
 def render_terminal_output(text: str) -> str:
@@ -113,9 +113,9 @@ def start_shell() -> pexpect.spawn:  # type: ignore
         encoding="utf-8",
         timeout=TIMEOUT,
     )
-    SHELL.expect(PROMPT)
+    SHELL.expect(PROMPT, timeout=TIMEOUT)
     SHELL.sendline("stty -icanon -echo")
-    SHELL.expect(PROMPT)
+    SHELL.expect(PROMPT, timeout=TIMEOUT)
     return SHELL
 
 
@@ -135,15 +135,15 @@ def _get_exit_code() -> int:
         return 0
     # First reset the prompt in case venv was sourced or other reasons.
     SHELL.sendline(f"export PS1={PROMPT}")
-    SHELL.expect(PROMPT)
+    SHELL.expect(PROMPT, timeout=0.2)
     # Reset echo also if it was enabled
     SHELL.sendline("stty -icanon -echo")
-    SHELL.expect(PROMPT)
+    SHELL.expect(PROMPT, timeout=0.2)
     SHELL.sendline("echo $?")
     before = ""
     while not _is_int(before):  # Consume all previous output
         try:
-            SHELL.expect(PROMPT)
+            SHELL.expect(PROMPT, timeout=0.2)
         except pexpect.TIMEOUT:
             print(f"Couldn't get exit code, before: {before}")
             raise
@@ -170,6 +170,7 @@ def initial_info() -> str:
 System: {uname_sysname}
 Machine: {uname_machine}
 Current working directory: {CWD}
+wcgw version: {importlib.metadata.version("wcgw")}
 """
 
 
@@ -215,7 +216,7 @@ def update_repl_prompt(command: str) -> bool:
 
 def get_cwd() -> str:
     SHELL.sendline("pwd")
-    SHELL.expect(PROMPT)
+    SHELL.expect(PROMPT, timeout=0.2)
     assert isinstance(SHELL.before, str)
     current_dir = render_terminal_output(SHELL.before).strip()
     return current_dir
@@ -342,13 +343,13 @@ def execute_bash(
         SHELL.expect(PROMPT)
         return "---\n\nFailure: user interrupted the execution", 0.0
 
-    wait = timeout_s or 5
+    wait = timeout_s or TIMEOUT
     index = SHELL.expect([PROMPT, pexpect.TIMEOUT], timeout=wait)
     if index == 1:
         BASH_STATE = "pending"
         text = SHELL.before or ""
 
-        text = render_terminal_output(text)
+        text = render_terminal_output(text[-100_000:])
         tokens = enc.encode(text)
 
         if max_tokens and len(tokens) >= max_tokens:
@@ -855,7 +856,7 @@ def get_tool_output(
         if imgBs64:
             console.print("Captured screenshot")
             outputs.append(ImageData(media_type="image/png", data=imgBs64))
-            if not IS_IN_DOCKER:
+            if not IS_IN_DOCKER and isinstance(arg, GetScreenInfo):
                 try:
                     # At this point we should go into the docker env
                     res, _ = execute_bash(
@@ -989,7 +990,7 @@ def read_file(readfile: ReadFile, max_tokens: Optional[int]) -> str:
 
     else:
         return_code, content, stderr = command_run(
-            f"cat {readfile.file_path}",
+            f"cat {readfile.file_path}", timeout=TIMEOUT
         )
         if return_code != 0:
             raise Exception(

wcgw/types_.py

@@ -60,12 +60,12 @@ class GetScreenInfo(BaseModel):
 
 class ScreenShot(BaseModel):
     type: Literal["ScreenShot"]
-    docker_image_id: str
 
 
 class MouseMove(BaseModel):
     x: int
     y: int
+    do_left_click_on_move: bool
     type: Literal["MouseMove"]
 
 

{wcgw-1.5.0.dist-info → wcgw-1.5.2.dist-info}/METADATA

@@ -1,13 +1,13 @@
 Metadata-Version: 2.3
 Name: wcgw
-Version: 1.5.0
+Version: 1.5.2
 Summary: What could go wrong giving full shell access to chatgpt?
 Project-URL: Homepage, https://github.com/rusiaaman/wcgw
 Author-email: Aman Rusia <gapypi@arcfu.com>
 Requires-Python: <3.13,>=3.11
 Requires-Dist: anthropic>=0.39.0
 Requires-Dist: fastapi>=0.115.0
-Requires-Dist: mcp>=1.0.0
+Requires-Dist: mcp
 Requires-Dist: mypy>=1.11.2
 Requires-Dist: nltk>=3.9.1
 Requires-Dist: openai>=1.46.0
@@ -29,7 +29,9 @@ Description-Content-Type: text/markdown
 
 # Shell and Coding agent on Chatgpt and Claude desktop apps
 
-A custom gpt on chatgpt web/desktop apps to interact with your local shell, edit files, run code, etc.
+- An MCP server on claude desktop for autonomous shell, coding and desktop control agent.
+- A custom gpt on chatgpt web/desktop apps to interact with your local shell, edit files, run code, etc.
+
 
 [![Tests](https://github.com/rusiaaman/wcgw/actions/workflows/python-tests.yml/badge.svg?branch=main)](https://github.com/rusiaaman/wcgw/actions/workflows/python-tests.yml)
 [![Build](https://github.com/rusiaaman/wcgw/actions/workflows/python-publish.yml/badge.svg)](https://github.com/rusiaaman/wcgw/actions/workflows/python-publish.yml)
@@ -40,6 +42,7 @@ A custom gpt on chatgpt web/desktop apps to interact with your local shell, edit
 ### 🚀 Highlights
 
 - ⚡ **Full Shell Access**: No restrictions, complete control.
+- ⚡ **Desktop control on Claude**: Screen capture, mouse control, keyboard control on claude desktop (on mac with docker linux)
 - ⚡ **Create, Execute, Iterate**: Ask the gpt to keep running compiler checks till all errors are fixed, or ask it to keep checking for the status of a long running command till it's done.
 - ⚡ **Interactive Command Handling**: Supports interactive commands using arrow keys, interrupt, and ansi escape sequences.
 - ⚡ **REPL support**: [beta] Supports python/node and other REPL execution.

{wcgw-1.5.0.dist-info → wcgw-1.5.2.dist-info}/RECORD

@@ -1,22 +1,22 @@
 wcgw/__init__.py,sha256=9K2QW7QuSLhMTVbKbBYd9UUp-ZyrfBrxcjuD_xk458k,118
-wcgw/types_.py,sha256=STKGeVdQNK_k8mcmAXpKZJY9YUpU7-mbJcFpXfzBuys,1732
+wcgw/types_.py,sha256=EUhSD8eQAJgoLb3Q4H5n57eged6t0ayVW0CLgFvs9jQ,1739
 wcgw/client/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 wcgw/client/__main__.py,sha256=wcCrL4PjG51r5wVKqJhcoJPTLfHW0wNbD31DrUN0MWI,28
-wcgw/client/anthropic_client.py,sha256=owR-nIxQVGgw_ned8JOQ-QmmCBQvZSqgD08kYpU_Rbg,17730
+wcgw/client/anthropic_client.py,sha256=_vFOraYHGjjMZKqdVOgH5CY6mUG0OQRYDH3img2LPEg,17325
 wcgw/client/cli.py,sha256=Oja42CHkVO8puqOXflko9NeephYCMa85aBmQTEjBZtI,932
 wcgw/client/common.py,sha256=grH-yV_4tnTQZ29xExn4YicGLxEq98z-HkEZwH0ReSg,1410
-wcgw/client/computer_use.py,sha256=jZCw4p7qfHEmhv9eqc1zyPyLTUyb54Ipw5UnRqzRWqc,14399
+wcgw/client/computer_use.py,sha256=1Sev4HhHix-FWNb77l5nJd2EXe906AkAiaDUNRLK-Zo,14605
 wcgw/client/diff-instructions.txt,sha256=s5AJKG23JsjwRYhFZFQVvwDpF67vElawrmdXwvukR1A,1683
 wcgw/client/openai_client.py,sha256=L61ajFVQW2QPS3C0n1YsjgF4vQKfMIZHmp6iFBHutX8,17748
 wcgw/client/openai_utils.py,sha256=YNwCsA-Wqq7jWrxP0rfQmBTb1dI0s7dWXzQqyTzOZT4,2629
-wcgw/client/sys_utils.py,sha256=7_7o1Au33OkZUsW5nKW55xW_YRYZlvUpY6tHVLIILm8,1254
-wcgw/client/tools.py,sha256=PbS1YBELqibShPrpKCGq3KeXLS4t6AhTTkB5pQMgX-g,32476
-wcgw/client/mcp_server/Readme.md,sha256=dK9NqFQ0_tblLOgPI6DjWrSYs2UWKEJkvgnLW3T-EZk,1944
+wcgw/client/sys_utils.py,sha256=GajPntKhaTUMn6EOmopENWZNR2G_BJyuVbuot0x6veI,1376
+wcgw/client/tools.py,sha256=d7Fni7JU3aOh2vXBAw5k5rsxkdQVcxoxc5vipvEsA2g,32680
+wcgw/client/mcp_server/Readme.md,sha256=1hNZtqltsORug7OzUjjoK5O8q5s9-Y3S0_rlzT-Wfg4,2033
 wcgw/client/mcp_server/__init__.py,sha256=cQ7PUrEmXUpio8x0SEoGWP5hCRPd7z2bAkNCbYbtTys,236
-wcgw/client/mcp_server/server.py,sha256=EzO4VNuSfcIMTue3sCsCqTTMku_AmsgJctBuPQfEQ8w,10160
+wcgw/client/mcp_server/server.py,sha256=kKtPlPUIerC3eWB10IfNK_erLWGIi-jU0MP99j2d4bc,9890
 wcgw/relay/serve.py,sha256=RUcUeyL4Xt0EEo12Ul6VQjb4tRle4uIdsa85v7XXxEw,8771
 wcgw/relay/static/privacy.txt,sha256=s9qBdbx2SexCpC_z33sg16TptmAwDEehMCLz4L50JLc,529
-wcgw-1.5.0.dist-info/METADATA,sha256=_bTBKk1cMtbqt7dgM-wge3LuNCTe4s1CDzdSyR7P3WU,6292
-wcgw-1.5.0.dist-info/WHEEL,sha256=C2FUgwZgiLbznR-k0b_5k3Ai_1aASOXDss3lzCUsUug,87
-wcgw-1.5.0.dist-info/entry_points.txt,sha256=eKo1omwbAggWlQ0l7GKoR7uV1-j16nk9tK0BhC2Oz_E,120
-wcgw-1.5.0.dist-info/RECORD,,
+wcgw-1.5.2.dist-info/METADATA,sha256=215jNmTFbkeETl2JrncCWwgEpUJU9H3MaE1pMHQEUyQ,6508
+wcgw-1.5.2.dist-info/WHEEL,sha256=C2FUgwZgiLbznR-k0b_5k3Ai_1aASOXDss3lzCUsUug,87
+wcgw-1.5.2.dist-info/entry_points.txt,sha256=eKo1omwbAggWlQ0l7GKoR7uV1-j16nk9tK0BhC2Oz_E,120
+wcgw-1.5.2.dist-info/RECORD,,