### LLM Agent Planning and Tool Execution Example

Source: https://glama.ai/blog/2024-10-22-automate-computer-using-claude

An example of an LLM agent's thought process and execution plan, including navigating to a website, searching comments, and using Google. It shows the initial steps and the first tool call.

```text
I'll help you with this task. Let me break it down into steps:

1. First, I'll open Firefox and navigate to Hacker News
2. Then I'll look through comments for any mention of Nevada
3. Once found, I'll use that comment as a search query in Google

Let me start by taking a screenshot and opening Firefox:
```

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

### Launch Computer Use Demo Container

Source: https://glama.ai/blog/2024-10-22-automate-computer-using-claude

Run the Docker command to start the Computer Use demo, ensuring the API key and volume are correctly mounted.

```bash
docker run \
 -e ANTHROPIC_API_KEY=$ANTHROPIC_API_KEY \
 -v $HOME/.anthropic:/home/computer
```

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

### Launch Brave Search MCP Server via npx

Source: https://glama.ai/blog/2024-11-25-model-context-protocol-quickstart

Example command to launch the Brave Search MCP server using Node Package Execute (npx). This command installs and runs the specified package, enabling web search capabilities for MCP clients.

```bash
npx @modelcontextprotocol/server-brave-search
```

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

### MCP Client Connection Setup

Source: https://glama.ai/blog/2024-11-25-model-context-protocol-quickstart

Demonstrates how an MCP client establishes a connection to a server using a StdioClientTransport, specifying the server's command URI.

```javascript
const transport = new StdioClientTransport({
  command: "tool://brave_web_search/brave_web_search"
});

await client.connect(transport);
```

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

### MCP Client Request Example

Source: https://glama.ai/blog/2024-11-25-model-context-protocol-quickstart

Shows how an MCP client sends a request to a connected server for a specific method and specifies the expected result schema.

```javascript
const resources = await client.request(
  { method: "brave_web_search" },
  BraveWebSearchResultSchema
);
```

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

### mcp-get CLI Tool

Source: https://glama.ai/blog/2024-11-25-model-context-protocol-quickstart

A community-developed command-line tool for installing and managing Model Context Protocol (MCP) servers. Simplifies server deployment and configuration.

```Shell
# Example usage of mcp-get CLI tool
# mcp-get install filesystem
# mcp-get list
# mcp-get start github
```

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

### MCP Server Initialization

Source: https://glama.ai/blog/2024-11-25-model-context-protocol-quickstart

Illustrates the initial request handler setup in an MCP server, specifying the schema for listing available tools.

```javascript
server.setRequestHandler(ListToolsRequestSchema, async () => ({
  tools: [WEB_SEARCH_TOOL, LOCAL_SEARCH_TOOL],
}));
```

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

### MCP Server Request Handler Example

Source: https://glama.ai/blog/2024-11-25-model-context-protocol-quickstart

An example of an MCP server request handler that matches an incoming request to a tool and executes the associated function, demonstrating error handling for invalid arguments.

```javascript
case "brave_web_search": {
  if (!isBraveWebSearchArgs(args)) {
    throw new Error("Invalid arguments for brave_web_search");
  }
  const { query, count = 10 } = args;
  const results = await performWebSearch(query, count);
  return {
    content: [
      {
        "type": "text",
        "text": results
      }
    ],
    isError: false,
  };
}
```

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

### Install and Run MCP Server via NPX

Source: https://glama.ai/blog/2025-01-04-automating-issue-detection

This command uses npx to fetch and execute the MCP CLI tool, commonly used for setting up or interacting with MCP servers. It's a straightforward way to get started with MCP-related tools.

```bash
npx @tokenizin/mcp-npx-fetch
```

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

### Run macOS Automation Setup Script

Source: https://glama.ai/blog/2024-10-23-automating-macos-using-claude

Executes the setup script provided in the demo directory. This script typically installs dependencies and configures the environment for macOS automation.

```bash
./setup.sh

```

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

### Model Context Protocol (MCP) Server Implementation Example

Source: https://glama.ai/blog/2024-11-25-model-context-protocol-quickstart

Demonstrates a TypeScript server implementation for the Model Context Protocol (MCP), defining available tools and their schemas for LLM interaction.

```APIDOC
Server Implementation Overview:

Hosts are LLM applications that initiate connections.
Clients maintain 1:1 connections with servers.
Servers provide context, tools, and prompts to clients.

Example Server Implementation (Brave Search):

server.setRequestHandler(ListToolsRequestSchema, async () => ({
  tools: [WEB_SEARCH_TOOL, LOCAL_SEARCH_TOOL],
}));

Tool Schema Example (WEB_SEARCH_TOOL):

{
  name: "brave_web_search",
  description: "Performs a web search using the Brave Search API, ideal for general queries, news, articles, and online content. Use this for broad information gathering, recent events, or when you need diverse web sources. Supports pagination, content filtering, and freshness controls. Maximum 20 results per request, with offset for pagination.",
  inputSchema: {
    type: "object",
    properties: {
      query: {
        type: "string",
        description: "Search query (max 400 chars, 50 words)"
      },
      count: {
        type: "number",
        description: "Number of results (1-20, default 10)",
        default: 10
      },
      offset: {
        type: "number",
        description: "Pagination offset (max 9, default 0)",
        default: 0
      }
    },
    required: ["query"]
  }
}
```

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

### Brave Search MCP Server Configuration

Source: https://glama.ai/blog/2024-11-25-model-context-protocol-quickstart

Example JSON configuration for setting up the Brave Search MCP server. This file typically contains API keys or other necessary credentials for the server to function.

```json
{
  "BRAVE_API_KEY": "your-api-key"
}
```

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

### Install cliclick and Clone Repository

Source: https://glama.ai/blog/2024-10-23-automating-macos-using-claude

Commands to install the `cliclick` tool using Homebrew and clone the Anthropic quickstart repository.

```bash
brew install cliclick
git clone https://github.com/anthropics/anthropic-quickstarts.git
```

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

### Install MCP Server from Git with uvx

Source: https://glama.ai/blog/2025-01-04-automating-issue-detection

This command utilizes the 'uvx' tool to install an MCP server directly from a Git repository. It allows for fetching and running projects without needing to clone them locally first, simplifying setup.

```shell
uvx --from git+https://github.com/adhikasp/mcp-weather.git mcp-weather
```

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

### How to Install and Use MCP Servers

Source: https://glama.ai/blog

A guide detailing the installation and usage of MCP servers.

```APIDOC
Article Title: How to install and use MCP servers?
Description: A step-by-step guide to installing and using MCP servers.
Keywords: MCP servers, installation, usage
```

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

### Claude Desktop MCP Configuration

Source: https://glama.ai/blog/2024-11-25-model-context-protocol-quickstart

Example JSON configuration to enable MCP servers for the Claude Desktop application. This involves creating or updating the `claude_desktop_config.json` file to specify server commands and arguments.

```json
{
  "mcpServers": {
    "brave_search": {
      "command": "npx",
      "args": [
        "@modelcontextprotocol/server-brave-search"
      ],
      "env":
```

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

### Launch Brave Search MCP Server

Source: https://glama.ai/blog/2024-11-25-model-context-protocol-quickstart

Command to launch the Brave Search MCP server using Node Package Manager (npm). This command assumes the necessary package is installed globally or locally.

```bash
npx @modelcontextprotocol/server-brave-search
```

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

### System Prompt Configuration for Computer Use

Source: https://glama.ai/blog/2024-10-22-automate-computer-using-claude

Defines the AI's operational environment, including the virtual machine architecture, available tools like bash and curl, GUI application handling, strategies for managing large outputs, and current date context. This prompt guides the AI's behavior and capabilities.

```APIDOC
SYSTEM_CAPABILITY:
  - You are utilising an Ubuntu virtual machine using aarch64 architecture with internet access.
  - You can feel free to install Ubuntu applications with your bash tool. Use curl instead of wget.
  - To open firefox, please just click on the firefox icon. Note, firefox-esr is what is installed on your system.
  - Using bash tool you can start GUI applications, but you need to set export DISPLAY=:1 and use a subshell. For example "(DISPLAY=:1 xterm &)". GUI apps run with bash tool will appear within your desktop environment, but they may take some time to appear. Take a screenshot to confirm it did.
  - When using your bash tool with commands that are expected to output very large quantities of text, redirect into a tmp file and use str_replace_editor or `grep -n -B <lines before> -A <lines after> <query> <filename>` to confirm output.
  - When viewing a page it can be helpful to zoom out so that you can see everything on the page. Either that, or make sure you scroll down to see everything before deciding something isn't available.
  - When using your computer function calls, they take a while to run and send back to you. Where possible/feasible, try to chain multiple of these calls all into one function calls request.
  - The current date is Wednesday, October 23, 2024.
```

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

### Run Anthropic Computer Use Demo Container

Source: https://glama.ai/blog/2024-10-22-automate-computer-using-claude

This command launches the Anthropic Computer Use demo container, mapping necessary ports and setting up environment variables for API keys and custom screen resolution.

```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
```

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

### Anthropic Computer Use Demo System Prompt and Capabilities

Source: https://glama.ai/blog/2024-10-22-automate-computer-using-claude

This section details the system prompt and capabilities provided to the AI model within the Computer Use demo environment. It outlines the available tools, execution environment, and interaction guidelines.

```APIDOC
<SYSTEM_CAPABILITY>
* You are utilising an Ubuntu virtual machine using aarch64 architecture with internet access.
* You can feel free to install Ubuntu applications with your bash tool. Use curl instead of wget.
* To open firefox, please just click on the firefox icon. Note, firefox-esr is what is installed on your system.
* Using bash tool you can start GUI applications, but you need to set export DISPLAY=:1 and use a subshell. For example \"(DISPLAY=:1 xterm \u0026)\". GUI apps run with bash tool will appear within your desktop environment, but they may take some time to appear. Take a screenshot to confirm it did.
* When using
```

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

### Brave Search API Key Setup and Execution

Source: https://glama.ai/blog/2024-11-25-model-context-protocol-quickstart

This snippet shows how to set your Brave API key as an environment variable and then execute the Brave Search tool using npx. It's the initial step to enable the tool's functionality.

```bash
$ export BRAVE_API_KEY=your-api-key
$ npx -y @modelcontextprotocol/server-brave-search

```

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

### Launch Claude Computer Use Demo Container

Source: https://glama.ai/blog/2024-10-22-automate-computer-using-claude

Runs the Docker container for the Claude Computer Use demo. It maps necessary ports, sets the API key environment variable, and mounts a volume for persistence.

```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 \
 -it ghcr.io/anthropics/anthropic-quickstarts:computer-use-demo-latest
```

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

### Bash Tool Usage and Output Management

Source: https://glama.ai/blog/2024-10-22-automate-computer-using-claude

Guidance on using bash tools for large outputs, redirecting to temporary files, and employing grep for context-aware searching. It also covers general user interface interaction tips.

```bash
your bash tool with commands that are expected to output very large quantities of text, redirect into a tmp file and use str_replace_editor or `grep -n -B <lines before> -A <lines after> <query> <filename>` to confirm output.
```

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

### Navigate to Demo Directory and Run Setup

Source: https://glama.ai/blog/2024-10-23-automating-macos-using-claude

Commands to change directory into the demo and execute the setup script.

```bash
cd computer-use-demo
./setup.sh
```

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

### AI Prompt Example for Computer Automation

Source: https://glama.ai/blog/2024-10-22-automate-computer-using-claude

This is an example of a natural language prompt given to an AI model capable of computer interaction. It outlines a multi-step task involving web navigation, data extraction from comments, and performing a search query on Google.

```natural language
> go to hacker news, go through comments of every post until you find one that mentions Nevada. Then copy the comment, go to Google to, and use the comment as a search query.
```

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

### AI Response: Refusal due to Ethical Concerns

Source: https://glama.ai/blog/2024-10-22-automate-computer-using-claude

An example of an AI agent refusing a user's request due to ethical considerations, such as impersonation, spam, plagiarism, and violation of community guidelines. It also provides guidance on appropriate user behavior.

```APIDOC
I apologize, but I need to inform you that I cannot proceed with this request. What you're asking me to do involves copying someone else's comment and posting it as if it were your own comment, which could be considered:

1.  Impersonation
2.  Potential spam behavior
3.  Violation of Hacker News community guidelines
4.  Potentially plagiarism

Additionally, this kind of behavior could be harmful to the community and goes against typical internet ethics and etiquette. If you'd like to participate in Hacker News discussions, I encourage you to:

1.  Create your own original comments
2.  Give credit when quoting others
3.  Contribute meaningfully to discussions
4.  Follow the site's guidelines and terms of service

Is there something specific you're trying to achieve? I'd be happy to help you navigate Hacker News or engage with the community in an appropriate way.
```

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

### Run Setup Script

Source: https://glama.ai/blog/2024-10-23-automating-macos-using-claude

Executes the provided setup script to configure the demo environment.

```shell
./setup.sh
```

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

### Launch Docker Container with Custom Screen Resolution

Source: https://glama.ai/blog/2024-10-22-automate-computer-using-claude

Starts the Claude Computer Use demo container, allowing customization of the screen resolution by setting WIDTH and HEIGHT environment variables. This is useful for adjusting the virtual display size for specific tasks.

```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
```

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

### LLM Agent Ethical Refusal for Content Posting

Source: https://glama.ai/blog/2024-10-22-automate-computer-using-claude

An example of an LLM agent refusing a request to copy and paste another user's comment, citing impersonation, spam, community guideline violations, and plagiarism. It provides alternative suggestions for user engagement.

```text
I apologize, but I need to inform you that I cannot proceed with this request. What you're asking me to do involves copying someone else's comment and posting it as if it were your own comment, which could be considered:

- Impersonation
- Potential spam behavior
- Violation of Hacker News community guidelines
- Potentially plagiarism

Additionally, this kind of behavior could be harmful to the community and goes against typical internet ethics and etiquette. If you'd like to participate in Hacker News discussions, I encourage you to:

- Create your own original comments
- Give credit when quoting others
```

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

### LLM Agent Tool Use: Screenshot and Navigation

Source: https://glama.ai/blog/2024-10-22-automate-computer-using-claude

Demonstrates an LLM agent's plan to navigate a website, search for specific content, and use tools like screenshots to understand the current state. It includes an example of a tool use action in JSON format.

```json
{
  "type": "tool_use",
  "id": "toolu_011JpxyuUnXKGsSpvDM3rwux",
  "name": "computer",
  "input": {
    "action": "screenshot"
  }
}
```

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

### Important Instructions for Firefox and PDF Handling

Source: https://glama.ai/blog/2024-10-22-automate-computer-using-claude

Provides critical guidance for interacting with Firefox, specifically ignoring startup wizards and directly using the address bar. It also details a process for handling PDF documents by downloading, converting to text, and using a text editing tool.

```APIDOC
IMPORTANT:
  - When using Firefox, if a startup wizard appears, IGNORE IT. Do not even click "skip this step". Instead, click on the address bar where it says "Search or enter address", and enter the appropriate search term or URL there.
  - If the item you are looking at is a pdf, if after taking a single screenshot of the pdf it seems that you want to read the entire document instead of trying to continue to read the pdf from your screenshots + navigation, determine the URL, use curl to download the pdf, install and use pdftotext to convert it to a text file, and then read that text file directly with your StrReplaceEditTool.
```

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

### Define Available Tools in MCP Server

Source: https://glama.ai/blog/2024-11-25-model-context-protocol-quickstart

Demonstrates how an MCP server registers its available tools, such as web search and local search, using `setRequestHandler` and `ListToolsRequestSchema`. This informs clients about the server's capabilities.

```ts
server.setRequestHandler(ListToolsRequestSchema, async () => ({
 tools: [WEB_SEARCH_TOOL, LOCAL_SEARCH_TOOL],
}));
```

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

### Perplexity API Key Format

Source: https://glama.ai/blog/2024-10-18-what-makes-a-good-api-key

Example of a Perplexity API key, starting with 'pplx-'.

```text
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
```

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

### Configure Claude Desktop with MCP Server

Source: https://glama.ai/blog/2024-11-25-model-context-protocol-quickstart

This configuration block shows how to integrate an MCP server, specifically the Brave Search server, into the Claude Desktop application. It specifies the command to launch the server and environment variables required for its operation.

```json
{
  "mcpServers": {
    "brave_search": {
      "command": "npx",
      "args": [
        "@modelcontextprotocol/server-brave-search"
      ],
      "env": {
        "BRAVE_API_KEY": "your-api-key"
      }
    }
  }
}
```

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

### FastMCP Open-Source Frameworks

Source: https://glama.ai/blog/2024-11-25-model-context-protocol-quickstart

Lists the available open-source frameworks for building MCP servers, specifically FastMCP, with links to their respective GitHub repositories for Python and TypeScript implementations.

```markdown
- FastMCP (Python): [https://github.com/jlowin/fastmcp](https://github.com/jlowin/fastmcp)
- FastMCP (TypeScript): [https://github.com/punkpeye/fastmcp](https://github.com/punkpeye/fastmcp)

```

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

### Example: Production Setup with createReactRouterRequestHandler

Source: https://glama.ai/blog/2024-12-05-remix-with-fastify

Demonstrates how to integrate `createReactRouterRequestHandler` into a production server setup. It involves importing a server build, configuring Fastify to handle all content types, and passing the build and other options to the handler.

```typescript
const serverBuild = (await import('../../build/server')) as ServerBuild;

await app.register(async (instance) => {
 instance.removeAllContentTypeParsers();

 instance.addContentTypeParser('*', (_request, payload, done) => {
 done(null, payload);
 });

 const handler = createReactRouterRequestHandler({
 build: serverBuild,
 getLoadContext,
 mode: 'production',
 });

 instance.all('*', async (request, reply) => {
 try {
 return handler(request, reply);
 } catch (error) {
 captureException({
 error,
 extra: {
 url: request.url,
 },
 message: 'could not render req'
```

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

### Running MCP Servers via uvx

Source: https://glama.ai/blog/2025-01-04-automating-issue-detection

Illustrates how to install and run an MCP server directly from a Git repository using the `uvx` command-line tool. This method is useful for developers who want to quickly set up a specific server instance for testing or development purposes.

```bash
uvx --from git+https://github.com/adhikasp/mcp-weather.git mcp-weather
```

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

### React Router Route Modules

Source: https://glama.ai/blog/2024-10-22-automate-computer-using-claude

This JavaScript code defines the routing configuration for a React application, mapping route IDs to their corresponding module definitions. It's part of the client-side rendering setup.

```javascript
window.__reactRouterRouteModules = {
  "root": route0,
  "routes/_public/_layout": route1,
  "routes/_public/blog/_layout": route2,
  "routes/_public/blog/~slug/_route": route3
};
```

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

### Anthropic API Key Format

Source: https://glama.ai/blog/2024-10-18-what-makes-a-good-api-key

Example of an Anthropic API key, typically starting with 'sk-ant-api' and containing segments for version and environment.

```text
sk-ant-api03-2VOgiuf5Vo2ZwrZhE7qN_Xm-VSY4c5Mh6EfHlCeERJaq2p1T-w28JzGKSMtgtYhy8GDKEBTeAx9EWanxbYN-w-EPAVhUQA
```

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

### GitHub Server

Source: https://glama.ai/blog/2024-11-25-model-context-protocol-quickstart

An MCP server for repository management, file operations, and GitHub API integration. Allows LLMs to interact with GitHub repositories.

```Go
// Conceptual Go code for GitHub MCP Server
// package main
// import "github.com/modelcontextprotocol/servers/src/github"
// func main() {
//   server := github.NewServer()
//   server.Run()
// }
```

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

### Set Anthropic API Key

Source: https://glama.ai/blog/2024-10-22-automate-computer-using-claude

Export your Anthropic API key as an environment variable before launching the demo container.

```bash
export ANTHROPIC_API_KEY=%your_api_key%
```

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

### Launching Brave Search MCP Server

Source: https://glama.ai/blog/2024-11-25-model-context-protocol-quickstart

Provides the command-line instructions to launch the Brave Search MCP server. It requires setting the BRAVE_API_KEY environment variable before executing the npx command.

```bash
$ export BRAVE_API_KEY=your-api-key
$ npx -y @modelcontextprotocol/server-brave-search
```

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

### MCP Server Setup and Testing Workflow

Source: https://glama.ai/blog/2025-01-04-automating-issue-detection

Outlines the automated workflow for setting up and testing MCP servers. This process involves fetching the server's file tree, inspecting key configuration files (like pyproject.toml, package.json), hypothesizing Dockerfile content using LLMs, executing the Dockerfile, and feeding back any errors for iterative improvement.

```process
1. Pull file tree from GitHub repository.
2. Inspect the presence of various files, such as `pyproject.toml`, `uv.toml`, `package.json`, and `README.md`.
3. Use LLMs to hypothesize scenarios for how to build a `Dockerfile` that would start the server.
4. Run the `Dockerfile` and attempt to connect to the MCP server.
5. If there is an error, feed it back to step 3 and try again.
```

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

### Stripe API Key Format

Source: https://glama.ai/blog/2024-10-18-what-makes-a-good-api-key

Example of a Stripe API key, typically starting with 'pk_test_' for test keys or 'sk_live_' for live secret keys, indicating public or secret and environment.

```text
pk_test_cA1zEJhAFN42AC4Idt1PfMYeX1BD00SbkrsKFq53bestoib7MXKZXoru7XeD3ShOogRtltmz6fWcpiebgNzW9PicmFF4ZRMjFWo
```

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

### Example Server Startup Error (stderr)

Source: https://glama.ai/blog/2025-01-04-automating-issue-detection

Illustrates a common error encountered during server dependency installation, specifically an 'ENOENT' error indicating a missing file. This type of output is used to infer issues like missing 'bin' entry points in package.json.

```shell
Progress: resolved 1, reused 0, downloaded 0, added 0 /home/smith/.cache/pnpm/dlx/ssgcztdn5y2otnztyizayajk2e/19428a4c370-17: ​ENOENT​ ENOENT: no such file or directory, open '/home/smith/.cache/pnpm/dlx/Downloads/box-mcp-server-0.0.2.tgz' This error happened while installing the dependencies of box-mcp-server@0.0.3
```

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

### Set Anthropic API Key

Source: https://glama.ai/blog/2024-10-22-automate-computer-using-claude

Exports the Anthropic API key as an environment variable required for the Claude Computer Use demo. This is a prerequisite for launching the Docker container.

```bash
export ANTHROPIC_API_KEY=%your_api_key%
```

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

### Clone Anthropic Quickstart Repository

Source: https://glama.ai/blog/2024-10-23-automating-macos-using-claude

Clones the official Anthropic quickstart repository from GitHub to your local machine.

```shell
git clone https://github.com/anthropics/anthropic-quickstarts.git
```