### Install Dependencies and Run Example

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-best-practices/README.md

Installs necessary Python packages and Playwright browsers, sets the API key, and runs a sample command to open TextEdit and type 'hello world'. This task will control your mouse and keyboard.

```bash
python -m pip install -r requirements.txt && python -m playwright install chromium
echo "ANTHROPIC_API_KEY=<YOUR API KEY HERE>" > .env
python -m computer_use "open TextEdit and type hello world"
```

--------------------------------

### Clone Repository and Navigate

Source: https://github.com/anthropics/claude-quickstarts/blob/main/browser-use-demo/README.md

Clone the quickstart repository and navigate into the browser-use-demo directory.

```bash
git clone https://github.com/anthropics/claude-quickstarts.git
cd claude-quickstarts/browser-use-demo
```

--------------------------------

### Run Generated Application Setup

Source: https://github.com/anthropics/claude-quickstarts/blob/main/autonomous-coding/README.md

Execute the setup script generated by the autonomous agent to prepare the application environment.

```bash
cd generations/my_project

# Run the setup script created by the agent
./init.sh
```

--------------------------------

### Setup and Run Docker Container

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-demo/README.md

Execute the setup script, build the Docker image, and run the container. Mounts the local repository for development and exposes necessary ports. Ensure your ANTHROPIC_API_KEY is set.

```bash
./setup.sh  # configure venv, install development dependencies, and install pre-commit hooks
docker build . -t computer-use-demo:local  # manually build the docker image (optional)
export ANTHROPIC_API_KEY=%your_api_key%
docker run \
    -e ANTHROPIC_API_KEY=$ANTHROPIC_API_KEY \
    -v $(pwd)/computer_use_demo:/home/computeruse/computer_use_demo/ `# mount local python module for development` \
    -v $HOME/.anthropic:/home/computeruse/.anthropic \
    -p 5900:5900 \
    -p 8501:8501 \
    -p 6080:6080 \
    -p 8080:8080 \
    -it computer-use-demo:local  # can also use ghcr.io/anthropics/anthropic-quickstarts:computer-use-demo-latest
```

--------------------------------

### Setup and Run Computer-Use Demo

Source: https://github.com/anthropics/claude-quickstarts/blob/main/CLAUDE.md

Commands to initialize the environment, build the Docker image, and execute the container for the computer-use demo.

```bash
./setup.sh
```

```bash
docker build . -t computer-use-demo:local
```

```bash
docker run -e ANTHROPIC_API_KEY=$ANTHROPIC_API_KEY -v $(pwd)/computer_use_demo:/home/computeruse/computer_use_demo/ -v $HOME/.anthropic:/home/computeruse/.anthropic -p 5900:5900 -p 8501:8501 -p 6080:6080 -p 8080:8080 -it computer-use-demo:local
```

--------------------------------

### Run the development server

Source: https://github.com/anthropics/claude-quickstarts/blob/main/financial-data-analyst/README.md

Start the local development environment to preview the application.

```bash
npm run dev
```

--------------------------------

### Install dependencies

Source: https://github.com/anthropics/claude-quickstarts/blob/main/financial-data-analyst/README.md

Install the required Node.js packages for the project.

```bash
npm install
```

--------------------------------

### Run Localization Demo

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-best-practices/README.md

Starts the localization demo server using uvicorn, enabling hot-reloading for development.

```bash
python -m uvicorn dev_ui.localization_demo.server:app --reload --port 8001
```

--------------------------------

### Install Test Dependencies

Source: https://github.com/anthropics/claude-quickstarts/blob/main/browser-use-demo/tests/README.md

Install the necessary packages to run the test suite.

```bash
# Install test dependencies
pip install -r test-requirements.txt

# Or install with extras
pip install -e ".[test]"
```

--------------------------------

### Install Headless Chromium with uv

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-best-practices/README.md

Install the headless Chromium browser using uv for potentially faster installation. This command synchronizes packages and then installs playwright.

```bash
uv sync
uv run playwright install chromium
```

--------------------------------

### Configure Environment Variables

Source: https://github.com/anthropics/claude-quickstarts/blob/main/browser-use-demo/README.md

Copy the example environment file and add your Anthropic API key.

```bash
cp .env.example .env
# Edit .env file and add your ANTHROPIC_API_KEY
```

--------------------------------

### Initial Project Orientation Commands

Source: https://github.com/anthropics/claude-quickstarts/blob/main/autonomous-coding/prompts/coding_prompt.md

Use these bash commands to understand the project structure, requirements, and history before starting development.

```bash
pwd
ls -la
cat app_spec.txt
cat feature_list.json | head -50
cat claude-progress.txt
git log --oneline -20
cat feature_list.json | grep '"passes": false' | wc -l
```

--------------------------------

### Install Pre-commit Hooks

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-demo/CONTRIBUTING.md

Installs pre-commit hooks to automate checks before commits.

```bash
pre-commit install
```

--------------------------------

### Run Docker Compose

Source: https://github.com/anthropics/claude-quickstarts/blob/main/browser-use-demo/README.md

Start the browser automation demo using Docker Compose. Use --watch for development with auto-sync.

```bash
# For production use:
docker-compose up --build

# For development with file watching (auto-sync changes):
docker-compose up --build --watch
```

--------------------------------

### Create and Run an Agent with Local and MCP Tools

Source: https://github.com/anthropics/claude-quickstarts/blob/main/agents/README.md

Instantiate an agent with a name, system prompt, local tools, and optionally configure MCP servers for extended capabilities. This example shows how to integrate the `ThinkTool` and set up standard input/output communication with an MCP server.

```python
from agents.agent import Agent
from agents.tools.think import ThinkTool

# Create an agent with both local tools and MCP server tools
agent = Agent(
    name="MyAgent",
    system="You are a helpful assistant.",
    tools=[ThinkTool()],  # Local tools
    mcp_servers=[
        {
            "type": "stdio",
            "command": "python",
            "args": ["-m", "mcp_server"],
        },
    ]
)

# Run the agent
response = agent.run("What should I consider when buying a new laptop?")
```

--------------------------------

### Install Development Dependencies

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-demo/CONTRIBUTING.md

Installs necessary packages for development using pip.

```bash
pip install -r dev-requirements.txt
```

--------------------------------

### Run Tool Panel Server

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-best-practices/README.md

Starts the tool panel server for manual tool execution. Open the provided URL in your browser to access the tool panel.

```bash
python -m uvicorn dev_ui.tool_panel.server:app --reload
# open http://127.0.0.1:8000
```

--------------------------------

### Install Claude Code and SDK

Source: https://github.com/anthropics/claude-quickstarts/blob/main/autonomous-coding/README.md

Install the latest versions of Claude Code CLI and the Claude Agent SDK using npm and pip.

```bash
# Install Claude Code CLI (latest version required)
npm install -g @anthropic-ai/claude-code

# Install Python dependencies
pip install -r requirements.txt
```

--------------------------------

### Install Headless Chromium

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-best-practices/README.md

Download the necessary headless Chromium browser for the browser tool. This is a one-time installation and requires approximately 150 MB.

```bash
python -m playwright install chromium
```

--------------------------------

### Browser Tool Setup

Source: https://github.com/anthropics/claude-quickstarts/blob/main/browser-use-demo/README.md

Initializes the BrowserTool and defines its parameters, including name, description, and input schema. This is a foundational step for integrating browser automation capabilities.

```python
browser_tool = BrowserTool()

def to_params(self):
    return {
        "name": "browser",
        "description": BROWSER_TOOL_DESCRIPTION,
        "input_schema": BROWSER_TOOL_INPUT_SCHEMA,
    }
```

--------------------------------

### Set Up Python Virtual Environment

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-best-practices/README.md

Creates and activates a Python virtual environment using a version of Python 3.11 or newer, then upgrades pip and installs project dependencies.

```bash
# Replace `python3.13` with whatever 3.11+ interpreter you installed.
python3.13 -m venv .venv
source .venv/bin/activate

python -m pip install --upgrade pip
python -m pip install -r requirements.txt
```

--------------------------------

### Manage Customer Support Agent Project

Source: https://github.com/anthropics/claude-quickstarts/blob/main/CLAUDE.md

Commands for installing dependencies, running the development server with various UI configurations, and building the project.

```bash
npm install
```

```bash
npm run dev
```

```bash
npm run dev:left
```

```bash
npm run dev:right
```

```bash
npm run dev:chat
```

```bash
npm run lint
```

```bash
npm run build
```

--------------------------------

### Verify Installations

Source: https://github.com/anthropics/claude-quickstarts/blob/main/autonomous-coding/README.md

Verify that Claude Code CLI and the Claude Code SDK are installed correctly.

```bash
claude --version  # Should be latest version
pip show claude-code-sdk  # Check SDK is installed
```

--------------------------------

### Manually Run Generated Application

Source: https://github.com/anthropics/claude-quickstarts/blob/main/autonomous-coding/README.md

Manually install dependencies and run the generated application, typically for Node.js projects.

```bash
cd generations/my_project

# Or manually (typical for Node.js apps):
npm install
npm run dev
```

--------------------------------

### Configure Computer Use with Environment Variables

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-best-practices/README.md

Demonstrates how to override configuration settings using environment variables for a specific command. This example disables browser tools and increases JPEG quality.

```bash
# one-off: turn off browser tools and bump JPEG quality
CU_ENABLE_BROWSER_USE_TOOLS=false CU_JPEG_QUALITY=85 \
  python -m computer_use "open Notes and type hello"
```

--------------------------------

### Set API Key

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-best-practices/README.md

Copies the example environment file and instructs to edit it, or alternatively, set the ANTHROPIC_API_KEY environment variable directly.

```bash
cp .env.example .env  # then edit .env
# or: export ANTHROPIC_API_KEY=...
```

--------------------------------

### Run Docker Container with Claude API Key

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-demo/README.md

This command starts the Docker container and sets the Anthropic API key. It maps several ports for accessing the demo app and VNC.

```bash
export ANTHROPIC_API_KEY=%your_api_key%
docker run \
    -e ANTHROPIC_API_KEY=$ANTHROPIC_API_KEY \
    -v $HOME/.anthropic:/home/computeruse/.anthropic \
    -p 5900:5900 \
    -p 8501:8501 \
    -p 6080:6080 \
    -p 8080:8080 \
    -it ghcr.io/anthropics/anthropic-quickstarts:computer-use-demo-latest
```

--------------------------------

### Initialize Project Environment

Source: https://github.com/anthropics/claude-quickstarts/blob/main/autonomous-coding/prompts/coding_prompt.md

Execute the init.sh script if it exists to set up the development environment. Ensure it has execute permissions first.

```bash
chmod +x init.sh
./init.sh
```

--------------------------------

### Initialize Agent with Server Tools

Source: https://github.com/anthropics/claude-quickstarts/blob/main/agents/agent_demo.ipynb

Initializes an agent named 'Server Tools Agent' with system instructions and a set of tools including web search, code execution, and a think tool. Configure the model with a specific name, max tokens, and temperature.

```python
server_agent = Agent(
    name="Server Tools Agent",
    system="""
You are a helpful assistant with access to:
1. Web search for finding current information
2. Code execution for running Python code
3. Think tool for complex reasoning

Use these tools effectively to answer questions that require current data or calculations.
""",
    tools=[think_tool, web_search_tool, code_execution_tool],
    config=ModelConfig(
        model="claude-sonnet-4-20250514",
        max_tokens=4096,
        temperature=0.7
    ),
    verbose=True
)
```

--------------------------------

### Use TOML Configuration File

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-best-practices/README.md

Illustrates how to use a TOML configuration file by pointing the CU_CONFIG environment variable to it. This allows for persistent configuration overrides.

```bash
# or keep a config.toml and point at it from .env
#   .env:
#     CU_CONFIG=config.toml
#   config.toml:
#     enable_advisor_tool = true
#     image_prune_strategy = "none"
#     extra_models = ["some-beta-model"]
```

--------------------------------

### Initialize agent with combined tools

Source: https://github.com/anthropics/claude-quickstarts/blob/main/agents/agent_demo.ipynb

Initializes the agent with a system prompt defining its capabilities and available tools. It includes standard tools and MCP servers for web search and calculation.

```python
# Create agent config 
system_prompt = """
You are a helpful assistant with access to:
1. Web search (brave_web_search, brave_local_search)
2. Mathematical calculator (calculate) 
3. A tool to think and reason (think)

Always use the most appropriate tool for each task.
"""


# Initialize agent with standard tools and MCP servers
agent = Agent(
    name="Multi-Tool Agent",
    system=system_prompt,
    tools=[think_tool], 
    mcp_servers=[brave_search_server, calculator_server], 
    config=ModelConfig(
        model="claude-3-7-sonnet-20250219", 
        max_tokens=4096,
        temperature=1.0
    ),
    verbose=True
)
```

--------------------------------

### Create and Activate Python Virtual Environment

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-demo/CONTRIBUTING.md

Sets up a Python virtual environment for development. Use 'source' on Unix-based systems and '.' on Windows.

```bash
python -m venv .venv
source .venv/bin/activate  # On Unix
# or
.venv\Scripts\activate  # On Windows
```

--------------------------------

### Create Anthropic Server Tools

Source: https://github.com/anthropics/claude-quickstarts/blob/main/agents/agent_demo.ipynb

Demonstrates the creation of Anthropic's native server tools for web search and code execution. Configure tool usage limits and block specific domains.

```python
# Create Anthropic server tools
web_search_tool = WebSearchServerTool(
    name="web_search",
    max_uses=5,  # Limit to 5 searches per request
    blocked_domains=["example.com"]  # Example of blocking specific domains
)

code_execution_tool = CodeExecutionServerTool()
```

--------------------------------

### Run Docker with Custom Screen Size

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-demo/README.md

Configures the demo application's screen resolution by setting WIDTH and HEIGHT environment variables. This is useful for testing different display sizes or adhering to recommended resolution limits.

```bash
docker run \
    -e ANTHROPIC_API_KEY=$ANTHROPIC_API_KEY \
    -v $HOME/.anthropic:/home/computeruse/.anthropic \
    -p 5900:5900 \
    -p 8501:8501 \
    -p 6080:6080 \
    -p 8080:8080 \
    -e WIDTH=1920 \
    -e HEIGHT=1080 \
    -it ghcr.io/anthropics/anthropic-quickstarts:computer-use-demo-latest
```

--------------------------------

### Display macOS Permission Prompt

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-best-practices/dev_ui/tool_panel/static/index.html

Conditionally displays a message and a dismiss button if required macOS permissions (screen recording or accessibility) are missing. The message guides the user on how to grant permissions.

```javascript
p => { if (p.screen_recording && p.accessibility) return; const el = document.getElementById('preflight'); const missing = []; if (!p.screen_recording) missing.push('Screen Recording'); if (!p.accessibility) missing.push('Accessibility'); el.textContent = 'Missing macOS permissions: ' + missing.join(', ') + '. Grant them in System Settings > Privacy & Security, then restart this server.'; const btn = document.createElement('button'); btn.className = 'btn'; btn.textContent = 'dismiss'; btn.onclick = () => { el.style.display = 'none'; }; el.appendChild(btn); el.style.display = 'block'; }
```

--------------------------------

### Configure tools and MCP servers

Source: https://github.com/anthropics/claude-quickstarts/blob/main/agents/agent_demo.ipynb

Sets up a standard Python tool (ThinkTool) and configures MCP servers for a calculator and Brave Search. The calculator uses a local Python script, while Brave Search requires an API key and uses npx to run the server.

```python
# Standard Python tool
think_tool = ThinkTool()

# Python MCP server
calculator_server_path = os.path.abspath(os.path.join(os.getcwd(), "tools/calculator_mcp.py"))
calculator_server = {
    "type": "stdio",
    "command": "python",
    "args": [calculator_server_path]
}
print(f"Calculator server configured: {'Yes' if calculator_server else 'No'}")


# Brave MCP server written in TypeScript

brave_api_key = os.environ.get("BRAVE_API_KEY_BASE_DATA", "")
print(f"Brave API key available: {'Yes' if brave_api_key else 'No'}")
brave_search_server = {
    "type": "stdio",
    "command": "npx",
    "args": ["-y", "@modelcontextprotocol/server-brave-search"],
    "env": {
        "BRAVE_API_KEY": brave_api_key, 
        "PATH": f"{os.path.dirname('npx')}:" + os.environ.get("PATH", "")
    }
}
print(f"Brave search server configured: {'Yes' if brave_search_server else 'No'}")
```

--------------------------------

### Run Tests

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-best-practices/README.md

Execute the test suite for the computer use best practices.

```bash
python -m pytest
```

--------------------------------

### Initialize UI and Fetch Model Data

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-best-practices/dev_ui/localization_demo/static/index.html

Fetches model metadata, populates dropdowns for models and thinking efforts, and sets up the default prompt and tool definitions. This code runs when the page loads.

```javascript
const $ = id => document.getElementById(id);
let effortSupport = new Set();
let defaultPrompt = '';

async function populate() {
    const r = await fetch('/meta').then(r => r.json());
    effortSupport = new Set(r.effort_supported_models);
    defaultPrompt = r.default_prompt_template;

    for (const m of r.models) {
        const o = document.createElement('option');
        o.value = o.textContent = m;
        if (m === r.default_model) o.selected = true;
        $('model').appendChild(o);
    }

    for (const t of r.thinking_efforts) {
        const o = document.createElement('option');
        o.value = o.textContent = t;
        $('effort').appendChild(o);
    }

    $('prompt').value = defaultPrompt;
    $('tooldef').textContent = JSON.stringify(r.point_at_tool, null, 2);

    syncEffort();
}

populate();
```

--------------------------------

### Clone the repository

Source: https://github.com/anthropics/claude-quickstarts/blob/main/financial-data-analyst/README.md

Download the project source code from the official repository.

```bash
git clone https://github.com/anthropics/anthropic-quickstarts.git
cd anthropic-quickstarts/financial-data-analyst
```

--------------------------------

### NPM Scripts for App Configurations

Source: https://github.com/anthropics/claude-quickstarts/blob/main/customer-support-agent/README.md

Use these scripts to run or build the application with specific sidebar configurations. They set environment variables to tailor the application's layout for development or production.

```bash
npm run dev: Runs the full app with both sidebars (default)
```

```bash
npm run build: Builds the full app with both sidebars (default)
```

```bash
npm run dev:full: Same as npm run dev
```

```bash
npm run dev:left: Runs the app with only the left sidebar
```

```bash
npm run dev:right: Runs the app with only the right sidebar
```

```bash
npm run dev:chat: Runs the app with just the chat area (no sidebars)
```

```bash
npm run build:full: Same as npm run build
```

```bash
npm run build:left: Builds the app with only the left sidebar
```

```bash
npm run build:right: Builds the app with only the right sidebar
```

```bash
npm run build:chat: Builds the app with just the chat area (no sidebars)
```

--------------------------------

### Run Demo with Limited Iterations

Source: https://github.com/anthropics/claude-quickstarts/blob/main/autonomous-coding/README.md

Run the autonomous coding agent demo with a maximum number of iterations for testing purposes.

```bash
python autonomous_agent_demo.py --project-dir ./my_project --max-iterations 3
```

--------------------------------

### Manage Financial Data Analyst Project

Source: https://github.com/anthropics/claude-quickstarts/blob/main/CLAUDE.md

Standard lifecycle commands for the Financial Data Analyst project.

```bash
npm install
```

```bash
npm run dev
```

```bash
npm run lint
```

```bash
npm run build
```

--------------------------------

### Build and Run Docker with Vertex AI Authentication

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-demo/README.md

Builds the Docker image and runs it using Google Cloud credentials for Vertex AI. It authenticates using Application Default Credentials and requires specifying the Vertex AI region and project ID.

```bash
docker build . -t computer-use-demo
gcloud auth application-default login
export VERTEX_REGION=%your_vertex_region%
export VERTEX_PROJECT_ID=%your_vertex_project_id%
docker run \
    -e API_PROVIDER=vertex \
    -e CLOUD_ML_REGION=$VERTEX_REGION \
    -e ANTHROPIC_VERTEX_PROJECT_ID=$VERTEX_PROJECT_ID \
    -v $HOME/.config/gcloud/application_default_credentials.json:/home/computeruse/.config/gcloud/application_default_credentials.json \
    -p 5900:5900 \
    -p 8501:8501 \
    -p 6080:6080 \
    -p 8080:8080 \
    -it computer-use-demo
```

--------------------------------

### Run the Agent

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-best-practices/README.md

Executes the computer-use agent to perform a task, such as opening TextEdit and typing a message.

```bash
python -m computer_use "open TextEdit and type hello world"
```

--------------------------------

### Test and Lint Computer-Use Demo

Source: https://github.com/anthropics/claude-quickstarts/blob/main/CLAUDE.md

Standard commands for maintaining code quality and running tests in the Python-based computer-use demo.

```bash
ruff check .
```

```bash
ruff format .
```

```bash
pyright
```

```bash
pytest
```

```bash
pytest tests/path_to_test.py::test_name -v
```

--------------------------------

### Configure environment variables

Source: https://github.com/anthropics/claude-quickstarts/blob/main/financial-data-analyst/README.md

Set up the API key required for Anthropic service integration.

```env
ANTHROPIC_API_KEY=your_api_key_here
```

--------------------------------

### Run Autonomous Agent Demo

Source: https://github.com/anthropics/claude-quickstarts/blob/main/autonomous-coding/README.md

Execute the autonomous coding agent demo script. Specify a project directory for the generated application.

```bash
python autonomous_agent_demo.py --project-dir ./my_project
```

--------------------------------

### Browser Automation Tools

Source: https://github.com/anthropics/claude-quickstarts/blob/main/autonomous-coding/prompts/coding_prompt.md

These are the available tools for UI testing. Use them to interact with the application like a human user and verify functionality and appearance.

```bash
puppeteer_navigate - Start browser and go to URL
puppeteer_screenshot - Capture screenshot
puppeteer_click - Click elements
puppeteer_fill - Fill form inputs
puppeteer_evaluate - Execute JavaScript (use sparingly, only for debugging)
```

--------------------------------

### Run Agent for Information Retrieval and Calculation

Source: https://github.com/anthropics/claude-quickstarts/blob/main/agents/agent_demo.ipynb

Uses the initialized agent to search for the current population of Tokyo and then execute Python code to calculate its population density per square kilometer, given its area.

```python
# Example 1: Use web search to find current information and code execution for analysis
server_agent.run("""
Search for the current population of Tokyo, Japan. 
Then write and execute Python code to calculate how many people that would be per square kilometer, 
given that Tokyo's area is approximately 2,194 square kilometers.
""")
```

--------------------------------

### Run Docker with AWS Profile Authentication

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-demo/README.md

Use this command to run the demo container when you want to leverage your existing AWS profile configuration. Ensure your AWS credentials file is correctly set up.

```bash
export AWS_PROFILE=<your_aws_profile>
docker run \
    -e API_PROVIDER=bedrock \
    -e AWS_PROFILE=$AWS_PROFILE \
    -e AWS_REGION=us-west-2 \
    -v $HOME/.aws:/home/computeruse/.aws \
    -v $HOME/.anthropic:/home/computeruse/.anthropic \
    -p 5900:5900 \
    -p 8501:8501 \
    -p 6080:6080 \
    -p 8080:8080 \
    -it ghcr.io/anthropics/anthropic-quickstarts:computer-use-demo-latest
```

--------------------------------

### JavaScript: Run Tool and Display Results

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-best-practices/dev_ui/tool_panel/static/index.html

Sends a POST request to the '/run' endpoint with tool and input data. Displays execution status, results, and any errors. Renders batch results or content blocks, including images.

```javascript
async function run(tool, input, delay) {
  const out = document.getElementById('out');
  let remaining = delay;
  const tick = () => {
    out.textContent = remaining > 0 ? `running in ${remaining.toFixed(1)}s ...` : 'running ...';
  };
  tick();
  const cd = delay > 0 ? setInterval(() => {
    remaining = Math.max(0, remaining - 0.1);
    tick();
  }, 100) : null;
  const r = await fetch('/run', {
    method: 'POST',
    headers: {'content-type': 'application/json'},
    body: JSON.stringify({tool, input, delay}),
  }).then(r => r.json());
  if (cd) clearInterval(cd);
  lastMeta = r.meta || null;
  const head = {ms: r.ms, error: r.error, meta: r.meta};
  out.textContent = JSON.stringify(head, null, 2);
  const imgDiv = document.getElementById('img');
  imgDiv.innerHTML = '';
  const batch = (r.meta && r.meta.batch_results) || null;
  if (batch && Array.isArray(batch)) {
    const ol = document.createElement('ol');
    ol.className = 'batch';
    for (const step of batch) {
      const li = document.createElement('li');
      if (step.error) li.className = 'err';
      const stepText = renderBlocks(step.content || [], li);
      const pre = document.createElement('pre');
      pre.textContent = (step.error ? 'error: ' + step.error + '\n' : '') + stepText;
      li.prepend(pre);
      ol.appendChild(li);
    }
    imgDiv.appendChild(ol);
  } else {
    const text = renderBlocks(r.content || [], imgDiv);
    if (text) out.textContent += '\n' + text;
  }
}
```

--------------------------------

### Pin Browser Viewport via Environment Variable

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-best-practices/README.md

Sets a specific browser viewport size using an environment variable. The string value is automatically coerced into a tuple.

```bash
# pin a viewport via env var (string is coerced to a tuple)
CU_BROWSER_VIEWPORT=1280x720 python -m computer_use "..."
```

--------------------------------

### Run Docker with AWS Access Key Authentication

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-demo/README.md

Use this command to run the demo container with explicit AWS access key, secret key, and session token. This is an alternative to using an AWS profile.

```bash
export AWS_ACCESS_KEY_ID=%your_aws_access_key%
export AWS_SECRET_ACCESS_KEY=%your_aws_secret_access_key%
export AWS_SESSION_TOKEN=%your_aws_session_token%
docker run \
    -e API_PROVIDER=bedrock \
    -e AWS_ACCESS_KEY_ID=$AWS_ACCESS_KEY_ID \
    -e AWS_SECRET_ACCESS_KEY=$AWS_SECRET_ACCESS_KEY \
    -e AWS_SESSION_TOKEN=$AWS_SESSION_TOKEN \
    -e AWS_REGION=us-west-2 \
    -v $HOME/.anthropic:/home/computeruse/.anthropic \
    -p 5900:5900 \
    -p 8501:8501 \
    -p 6080:6080 \
    -p 8080:8080 \
    -it ghcr.io/anthropics/anthropic-quickstarts:computer-use-demo-latest
```

--------------------------------

### Configure Environment Variables

Source: https://github.com/anthropics/claude-quickstarts/blob/main/customer-support-agent/README.md

Define required API keys in a .env.local file. Note the 'B' prefix used for AWS variables.

```text
ANTHROPIC_API_KEY=your_anthropic_api_key
BAWS_ACCESS_KEY_ID=your_aws_access_key
BAWS_SECRET_ACCESS_KEY=your_aws_secret_key
```

--------------------------------

### Render Tool Cards from Fetch Data

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-best-practices/dev_ui/tool_panel/static/index.html

Fetches tool definitions from a '/tools' endpoint and dynamically renders interactive cards for each tool. Each card includes the tool's name, description, input fields, and a 'run' button with a delay setting.

```javascript
fetch('/tools').then(r => r.json()).then(tools => { const root = document.getElementById('tools'); for (const t of tools) { const card = document.createElement('div'); card.className = 'card'; card.innerHTML = `<h3>${t.name}</h3>` + `<details><summary>description</summary><p>${t.description}</p></details>`; for (const [k, spec] of Object.entries(t.input_schema.properties || {})) card.appendChild(fieldFor(k, spec)); const row = document.createElement('div'); row.className = 'row'; const delay = document.createElement('input'); delay.type = 'number'; delay.value = '3'; delay.title = 'delay (s)'; const btn = document.createElement('button'); btn.className = 'btn'; btn.textContent = 'run'; btn.onclick = () => run(t.name, collect(card), parseFloat(delay.value) || 0); row.append('delay:', delay, btn); card.appendChild(row); root.appendChild(card); } });
```

--------------------------------

### Handle Image File Setting

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-best-practices/dev_ui/localization_demo/static/index.html

Sets the selected image file for preview and display. It creates a DataTransfer object to set the file and updates the image source and filename display. Also handles adding a 'has-file' class to the drop zone.

```javascript
function setFile(file) {
    if (!file) return;
    const dt = new DataTransfer();
    dt.items.add(file);
    $('image').files = dt.files;
    $('preview').src = URL.createObjectURL(file);
    $('filename').textContent = `${file.name} (${(file.size/1024).toFixed(0)} KB)`;
    $('drop').classList.add('has-file');
}

$('image').addEventListener('change', e => setFile(e.target.files[0]));
```

--------------------------------

### Test Directory Structure

Source: https://github.com/anthropics/claude-quickstarts/blob/main/browser-use-demo/tests/README.md

Overview of the project's test file organization.

```text
tests/
├── conftest.py                      # Shared fixtures and mocks
├── test_message_renderer.py         # MessageRenderer class tests (~300 test cases)
├── test_streamlit_helpers.py        # Helper function tests (~150 test cases)
└── test_integration.py              # End-to-end integration tests (~50 test cases)
```

--------------------------------

### Run Ruff for Linting and Formatting

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-demo/CONTRIBUTING.md

Commands to check code for linting errors and format the code using Ruff.

```bash
ruff check .
ruff format .
```

--------------------------------

### Layout Styling for Computer Use Demo

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-demo/image/static_content/index.html

Defines the CSS for a full-screen split layout with two side-by-side iframes.

```css
body { margin: 0; padding: 0; overflow: hidden; } .container { display: flex; height: 100vh; width: 100vw; } .left { flex: 1; border: none; height: 100vh; } .right { flex: 2; border: none; height: 100vh; }
```

--------------------------------

### Open Screen Recording Permissions

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-best-practices/README.md

Opens the macOS System Settings pane for Screen Recording permissions. This is required for the agent to take screenshots.

```bash
open "x-apple.systempreferences:com.apple.preference.security?Privacy_ScreenCapture"
```

--------------------------------

### Run CI Test Pipeline

Source: https://github.com/anthropics/claude-quickstarts/blob/main/browser-use-demo/tests/README.md

Commands for executing tests and generating coverage reports in a CI environment.

```bash
# Install dependencies
pip install -e ".[test]"

# Run tests with coverage
pytest tests/ --cov=browser_tools_api_demo --cov-report=xml --cov-report=term

# Generate coverage badge
coverage-badge -o coverage.svg
```

--------------------------------

### View Trajectory with Streamlit

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-best-practices/README.md

Launches the trajectory viewer using Streamlit, which allows for inspecting past agent runs as chat transcripts with tool calls and results.

```bash
python -m streamlit run dev_ui/trajectory_viewer/app.py
```

--------------------------------

### Settings API

Source: https://github.com/anthropics/claude-quickstarts/blob/main/autonomous-coding/prompts/app_spec.txt

Endpoints for managing user and application settings.

```APIDOC
## GET /api/settings

### Description
Retrieves the current user settings.

### Method
GET

### Endpoint
/api/settings

## PUT /api/settings

### Description
Updates the current user settings.

### Method
PUT

### Endpoint
/api/settings

## GET /api/settings/custom-instructions

### Description
Retrieves custom instructions for the user.

### Method
GET

### Endpoint
/api/settings/custom-instructions

## PUT /api/settings/custom-instructions

### Description
Updates custom instructions for the user.

### Method
PUT

### Endpoint
/api/settings/custom-instructions
```

--------------------------------

### Commit Development Progress

Source: https://github.com/anthropics/claude-quickstarts/blob/main/autonomous-coding/prompts/coding_prompt.md

Create a descriptive git commit message detailing the implemented feature, verification steps, and changes made. Include specific details about the implementation and testing.

```bash
git add .
git commit -m "Implement [feature name] - verified end-to-end

- Added [specific changes]
- Tested with browser automation
- Updated feature_list.json: marked test #X as passing
- Screenshots in verification/ directory
"
```

--------------------------------

### Execute Test Suite

Source: https://github.com/anthropics/claude-quickstarts/blob/main/browser-use-demo/tests/README.md

Commands to run tests using pytest with various options for coverage and granularity.

```bash
pytest tests/
```

```bash
pytest tests/ --cov=browser_tools_api_demo --cov-report=html
# Open htmlcov/index.html to view coverage report
```

```bash
pytest tests/test_message_renderer.py -v
```

```bash
pytest tests/test_message_renderer.py::TestMessageRenderer -v
pytest tests/test_message_renderer.py::TestRenderMethod::test_render_string_message -v
```

```bash
# Run only integration tests
pytest -m integration

# Run tests excluding integration
pytest -m "not integration"

# Run async tests
pytest -m asyncio
```

--------------------------------

### Run Agent Query

Source: https://github.com/anthropics/claude-quickstarts/blob/main/agents/agent_demo.ipynb

Use the agent.run() method to execute a natural language query. The agent will determine the necessary tools and steps to fulfill the request.

```python
agent.run("What's the square root of the OKC population in 2022")
```

--------------------------------

### Open Accessibility Permissions

Source: https://github.com/anthropics/claude-quickstarts/blob/main/computer-use-best-practices/README.md

Opens the macOS System Settings pane for Accessibility permissions. This is required for the agent to control the mouse and keyboard.

```bash
open "x-apple.systempreferences:com.apple.preference.security?Privacy_Accessibility"
```

--------------------------------

### Import necessary libraries

Source: https://github.com/anthropics/claude-quickstarts/blob/main/agents/agent_demo.ipynb

Imports required libraries for agent functionality, including system path manipulation and specific tool modules. Applies nest_asyncio to handle nested asyncio loops.

```python
import os
import sys
import nest_asyncio
nest_asyncio.apply()

parent_dir = os.path.dirname(os.getcwd())
sys.path.insert(0, parent_dir)
from agents.agent import Agent, ModelConfig
from agents.tools.think import ThinkTool
from agents.tools.web_search import WebSearchServerTool
from agents.tools.code_execution import CodeExecutionServerTool
```