Tools, Resources & Prompts

The three types of capabilities your AI can access through MCP

The Three Types of MCP Capabilities

When you connect to an MCP server (like GitHub or Slack), your AI gets access to three types of capabilities:

Tools 🔧

Tools are actions your AI can perform - things that change or do something.

Examples:

• GitHub: Create an issue, merge a pull request, comment on code
• Slack: Send a message, create a channel, react to posts
• Dropbox: Upload a document, share a file, organize files in folders
• Email: Send an email, mark as read, move to folder

How your AI uses tools:

You: "Create a GitHub issue for the login bug"
AI: [Uses GitHub create_issue tool] "Issue #123 created: Fix login bug"

You: "Send a Slack message to the team"  
AI: [Uses Slack send_message tool] "Message sent to #general"

Resources 📄

Resources are information your AI can read - things that provide data.

Examples:

• GitHub: Repository contents, issue descriptions, commit history
• Slack: Message history, channel member lists, thread contents
• Dropbox: File contents, folder structures, file metadata
• Databases: Query results, table schemas, stored data

How your AI uses resources:

You: "What's in our latest pull request?"
AI: [Reads GitHub PR resource] "The PR adds user authentication with OAuth..."

You: "Summarize yesterday's Slack messages"
AI: [Reads Slack message history resource] "Main topics discussed: deployment, bug fixes..."

Prompts 💭

Prompts are pre-written instructions that help your AI use tools and resources effectively.

Examples:

• Code Review Prompt: “Review this pull request for security issues, performance, and code style”
• Bug Report Prompt: “Create a detailed bug report with steps to reproduce, expected vs actual behavior”
• Meeting Summary Prompt: “Summarize key decisions, action items, and next steps from this conversation”

How your AI uses prompts:

You: "Review this pull request"
AI: [Uses code review prompt] Systematically checks security, performance, style, and provides structured feedback

You: "Summarize today's standup"  
AI: [Uses meeting summary prompt] Creates organized summary with decisions, action items, blockers

How They Work Together

A typical AI workflow combines all three:

1. Read Resources

   AI reads current state (GitHub issues, Slack messages, file contents)

2. Use Prompts

   AI applies pre-written instructions for how to analyze or process the information

3. Execute Tools

   AI takes actions based on what it learned (create issues, send messages, update files)

Example Workflow:

You: "Prepare for our team standup"

1. AI reads resources: Recent GitHub commits, Slack messages, project files
2. AI uses prompt: "Standup preparation checklist"
3. AI uses tools: Creates summary document, posts reminder in Slack

Understanding MCP Server Capabilities

Each MCP server provides different combinations:

GitHub Server

• Tools: Create issues, comment, merge PRs, create branches
• Resources: Repository files, issue lists, commit history, PR details
• Prompts: Code review templates, issue creation guidelines

Slack Server

• Tools: Send messages, create channels, react to posts, set status
• Resources: Message history, channel lists, user profiles, thread contents
• Prompts: Message formatting, team communication guidelines

Dropbox Server

• Tools: Upload documents, search files, organize folders
• Resources: File contents, folder structures, sharing permissions
• Prompts: Document creation templates, organization standards

Why This Matters

Understanding these three types helps you:

• Ask better questions: “Read the latest GitHub issues” (resource) vs “Create a new issue” (tool)
• Set expectations: Know what your AI can read vs what it can change
• Troubleshoot: If something isn’t working, understand whether it’s a tool, resource, or prompt issue
• Explore capabilities: Ask “What tools are available?” to discover what actions your AI can take

Discovering Available Capabilities

Try these prompts to explore what’s available:

"What tools can you use?"
"What information can you access?"  
"What prompts are available for GitHub?"
"Show me all the Slack capabilities"

Your AI will list the tools, resources, and prompts available from your connected MCP servers, helping you understand what’s possible.