Learning Log #2: Claude and Its Capabilities
What I explored
What I explored
In this learning log, I wanted to understand Claude’s capabilities beyond the normal “ask a question, get an answer” chat interface. I had already used Claude casually, but I didn’t realize how deep the lore went until a friend mentioned MCP servers. That’s when I learned there are levels to how Claude can interact with tools, files, and external systems.
I followed Anthropic’s official module, Introduction to Model Context Protocol, to understand how MCP works and how it expands what Claude can do.
Key takeaways
What MCP is…
MCP stands for Model Context Protocol. It sorta acts as a bridge that lets AI tools like Claude connect to:
local files
databases
APIs
software tools
external data sources
Basically, it gives Claude controlled access to things it normally can’t see. I’ve seen videos of people giving Claude access to their system and having it clean up their desktop. Might be the better option than trying to download a file cleaner in the app store.
Why not just use Claude normally?
I asked that question a lot. It took me a bit to understand- especially since now Claude can access the web.
I learned that MCP servers give you:
more control
more customization
strict boundaries on what the AI can access
the ability to define exactly what Claude is allowed to read or do
Normally, if you ask an AI to read your files, it’ll say it can’t. It would be a bit concerning if it could. With MCP configured correctly, it can- safely.
Real‑world use cases
I found an article where someone configured an MCP server to help with OSINT research. OSINT takes time because you have to gather, filter, and process a lot of information. With MCP:
It can gather data
It can process it
It can help organize your report
I’ve mentioned it before- it doesn’t replace the analyst, but it speeds up the workflow.
Setting up my own MCP server!
I set up an MCP server on my MacBook, with Claude Desktop acting as the client. The nice part is that MCP servers can run over different protocols, so even though most people run the client and server on the same machine, you don’t haveto. There’s flexibility depending on your setup. I was thinking of creating one on my VM.
My Thoughts
I didn’t realize how powerful Claude could be until I learned about MCP servers. Before this, I was basically using Claude on “level one”- typing prompts and getting answers. Now I understand that Claude can be integrated into actual workflows, like a mini assistant.
It also made me think about how AI is evolving. The more control users have, the more useful AI becomes- but also the more important it is to set boundaries.
Link: AnthropicAIAI Learning Resources & Guides from Anthropic
