Block VI · Coding with agents · Lesson 33

Essential extensions, skills and MCP

You'll get to know the add-ons that most multiply what an agent can do: always up-to-date documentation, a real browser, access to GitHub, precise symbol-level editing and test-driven workflows.

Go to the course

What you'll see in this lesson

  1. Five add-ons that supercharge your agent

    An agent only knows what it learned in training and what it sees in your folder; add-ons give it eyes, hands and a method.

  2. Context7: up-to-date docs, not from memory

    Context7 gives the agent the current docs for the version you use, so it doesn't invent old or nonexistent functions.

  3. Playwright: let the agent test in a real browser

    With Playwright the agent opens your website, clicks, fills in forms and checks whether what it coded really works.

  4. GitHub MCP: issues, PRs and Actions from the chat

    GitHub's official server lets the agent read issues, open PRs or check a CI failure without you copying and pasting anything.

  5. Serena: edit by symbols, like an IDE

    Serena gives the agent IDE tools: find references, rename across the whole project or edit an entire function at once.

  6. Superpowers: a test-driven work method

    Superpowers is a set of skills that makes the agent ask, plan, write the test first and review before wrapping up.

  7. Less is more: every add-on takes up context

    Each MCP server adds its tool descriptions to the context; with too many, the agent spends more and chooses worse.

  8. Your kit by project type

    You don't need all five at once: combine two or three depending on what you're doing and add the rest when you need them.

  9. Summary: must-have add-ons

    With two or three well-chosen add-ons, the agent makes fewer mistakes, checks its work and follows a method.

Tools and concepts

Sign in with your Google account to access this lesson and the whole course.

Go to the courseSee the full syllabus

More lessons in this block