Welcome to Day 7! Yesterday we looked inside an agent and saw how it turns your words into actions.

Today we’re going a level deeper to understand how agents are able to do things. Without tools, an agent can only talk. With tools, an agent can actually GSD.

🧠 What You’ll Learn Today

• What “tools” mean in the context of agent systems
• How MCP tools give agents access to external data and services
• A real example: tools used inside GitLab Duo Agent Platform

🔧 What Are Agent Tools?

Tools are effectively the capabilities and permissions an agent can use.

Like a cross between an API call and access controls, a tool is the capability that defines what an agent can do:

• Searching a calendar
• Sending a message
• Editing a file
• Running a workflow
• Pulling data

🔌 MCP: A Universal Adapter for Tools

You’ll increasingly hear about MCP (Model Context Protocol) — a standardized way for agents to access tools, databases, APIs, and systems.

Instead of wiring everything together manually, MCP provides:

• A consistent format
• Clear permissions
• Predictable behavior
• A safe boundary between your agent and your data

MCP has brought standardization to agent access and actions, enabling interoperability between tools and contributing to the explosive growth of the AI tools market.

🛠️ Real Tools in Action: GitLab Duo Agent Platform

GitLab Duo Agent Platform is a great example of how tool libraries enable agents to perform complex, multi-step development tasks.

For those not familiar, GitLab is a platform for software development. With that context in mind, here’s a tiny slice of what agents can do in GitLab Duo Agent Platform:

• Search things (e.g., gitlab_blob_search, gitlab_issue_search, gitlab_documentation_search)
• Create things (create_issue, create_epic, create_merge_request)
• Edit things (edit_file, create_file_with_contents)

All of these capabilities (and dozens more) come directly from the built-in tool definitions in the platform.

That means a GitLab agent can:

• Find the right file Understand context
• Edit code
• Run tests to make sure the code works
• Raise issues
• Review merge requests
• Even analyze security vulnerabilities

Now imagine orchestrating multiple agents, with different tools, together.

🛠️ Hands-On Micro Experiment

Try this with Gemini or ChatGPT with tools enabled:

• Ask “what tools do you have access to?” This will provide an explanation of what these agents can do for you right now.
• Now ask, “how can I add more tools?” This will give you an idea of what you can do with a little bit of effort.

✍️ Quick Recap

• Tools are how agents act
• MCP standardizes tool access and makes integrations safer
• GitLab Duo Agent Platform is an example of tool-rich agent architecture

📬 Keep the Streak Going

Tomorrow: a special Friday missive!

👉 Subscribe at 12DaysofAgents.com

👉 Forward Day 7 to someone curious about how AI actually does things