Do I need a paid GitHub plan?

No. A free GitHub account with a personal access token that has repo scope is enough. You can test against your own public repositories.

Which LLM providers are supported?

The workshop ships with a provider abstraction that supports Fireworks, OpenRouter, Gemini, and OpenAI. You only need one API key to follow along.

Do I need Claude Desktop installed?

It helps, because you will wire the server into Claude Desktop, but the FastAPI inspector endpoint lets you test every tool without any MCP client installed.

Is this compatible with Cursor?

Yes. The server speaks stdio MCP, which Cursor and any other compliant MCP client understands. The config block is almost identical to Claude Desktop.

Premium course

Build an MCP server that actually reviews your pull requests

Name: Building MCP servers for GitHub PR review
Price: 49 USD
Availability: InStock

Stop copy pasting diffs into Claude. Build a Python MCP server that fetches your PR, runs structured LLM review, and posts inline comments back to GitHub, all callable from Claude Desktop or Cursor.

Enroll Preview curriculum

Still deciding? Ask first.

Message a mentor about fit, prerequisites, or where to start. Replies come on WhatsApp, usually within a day.

Curriculum fit, prerequisites, or where to start
Honest answer, no pressure to enroll

Engineers are learning here from

NVIDIAMICROSOFTGRABWISEPIPEDRIVEBOLTGLIA

Build a production Model Context Protocol server that fetches pull request diffs, runs LLM-powered code review, and posts structured review comments back to GitHub. Exposes MCP tools any client like Claude Desktop or Cursor can call.

Build an MCP server that reviews GitHub pull requests end to end.

What you'll ship

Real projects, not toy demos.

An MCP server that Claude Desktop and Cursor can spawn over stdio
A fetch_pr_diff tool that returns a unified diff plus changed files
A review_pr tool that feeds diffs to an LLM and returns structured comments
A typed ReviewComment schema anchored to file path and line number
A post_review_comments tool that writes the review back to GitHub
A FastAPI inspector endpoint for debugging the pipeline outside MCP clients

What you'll learn

You finish able to:

Stand up a Python MCP server over stdio that Claude Desktop and Cursor can spawn
Expose tools with typed input schemas using the official MCP SDK
Fetch pull request diffs with PyGithub and the GitHub diff media type
Chunk large diffs so they fit inside a reasonable LLM context budget
Design a structured ReviewComment schema with Pydantic and a severity taxonomy
Post review comments back to GitHub with idempotency and dedupe built in
Run the same business logic behind a FastAPI inspector for fast debugging

Curriculum

From an empty MCP server to a reviewer that writes back to GitHub.

01
MCP server skeleton
Stand up a Python MCP server over stdio and wire it into Claude Desktop
3 lessons
02
fetch_pr_diff tool
Fetch unified diffs from GitHub with proper authentication, rate limit handling, and chunking
3 lessons
03
review_pr tool
Feed the diff to an LLM, parse structured JSON, and classify feedback by severity
2 lessons
04
ReviewComment schema
Model review comments with Pydantic and anchor them to the right file and line
2 lessons
05
post_review_comments tool
Write the review back to GitHub with idempotent behavior and clean dedupe
2 lessons
06
FastAPI inspector UI
Add a debug UI that replays any PR and recaps the full build
2 lessons

Who it's for

Is this for you?

Python backend engineers

who have read the MCP spec but never built a server that talks to a real API

AI engineers

who want Claude Desktop to do useful work on their repos, not just answer questions

Platform engineers

who want to expose internal tools to LLM clients through a standard protocol

FAQ

Common questions.

Do I need a paid GitHub plan?
No. A free GitHub account with a personal access token that has repo scope is enough. You can test against your own public repositories.
Which LLM providers are supported?
The workshop ships with a provider abstraction that supports Fireworks, OpenRouter, Gemini, and OpenAI. You only need one API key to follow along.
Do I need Claude Desktop installed?
It helps, because you will wire the server into Claude Desktop, but the FastAPI inspector endpoint lets you test every tool without any MCP client installed.
Is this compatible with Cursor?
Yes. The server speaks stdio MCP, which Cursor and any other compliant MCP client understands. The config block is almost identical to Claude Desktop.

Pricing

Pick the path that fits.

Subscribe to Pro for every paid course, or buy just this one.

Subscribe to Pro

Recommended

Unlock this course and every paid course plus workshop replays. One subscription.

$35$16/mo · Billed annually

You save 54% with regional pricing

Every paid course unlocked
All workshop replays included
New releases as they ship
Cancel anytime

Buy this course

One-time purchase. Lifetime access to every lesson, exercise, and update.

$149$79one-time

You save 47% with regional pricing

Lifetime access to every lesson and future update
Hands-on projects you can show in interviews
Certificate when you finish

Still deciding? Ask Param a question

After this course:

MCP is how LLMs talk to your systems. Build one that does real work.

Enroll

Building MCP servers for GitHub PR review

$79 one-time

Enroll

Premium course

Build an MCP server that actually reviews your pull requests

Stop copy pasting diffs into Claude. Build a Python MCP server that fetches your PR, runs structured LLM review, and posts inline comments back to GitHub, all callable from Claude Desktop or Cursor.

Enroll Preview curriculum

Still deciding? Ask first.

Message a mentor about fit, prerequisites, or where to start. Replies come on WhatsApp, usually within a day.

Curriculum fit, prerequisites, or where to start
Honest answer, no pressure to enroll

Engineers are learning here from

NVIDIAMICROSOFTGRABWISEPIPEDRIVEBOLTGLIA

Build an MCP server that reviews GitHub pull requests end to end.

What you'll ship

Real projects, not toy demos.

An MCP server that Claude Desktop and Cursor can spawn over stdio
A fetch_pr_diff tool that returns a unified diff plus changed files
A review_pr tool that feeds diffs to an LLM and returns structured comments
A typed ReviewComment schema anchored to file path and line number
A post_review_comments tool that writes the review back to GitHub
A FastAPI inspector endpoint for debugging the pipeline outside MCP clients

What you'll learn

You finish able to:

Stand up a Python MCP server over stdio that Claude Desktop and Cursor can spawn
Expose tools with typed input schemas using the official MCP SDK
Fetch pull request diffs with PyGithub and the GitHub diff media type
Chunk large diffs so they fit inside a reasonable LLM context budget
Design a structured ReviewComment schema with Pydantic and a severity taxonomy
Post review comments back to GitHub with idempotency and dedupe built in
Run the same business logic behind a FastAPI inspector for fast debugging

Curriculum

From an empty MCP server to a reviewer that writes back to GitHub.

01
MCP server skeleton
Stand up a Python MCP server over stdio and wire it into Claude Desktop
3 lessons
02
fetch_pr_diff tool
Fetch unified diffs from GitHub with proper authentication, rate limit handling, and chunking
3 lessons
03
review_pr tool
Feed the diff to an LLM, parse structured JSON, and classify feedback by severity
2 lessons
04
ReviewComment schema
Model review comments with Pydantic and anchor them to the right file and line
2 lessons
05
post_review_comments tool
Write the review back to GitHub with idempotent behavior and clean dedupe
2 lessons
06
FastAPI inspector UI
Add a debug UI that replays any PR and recaps the full build
2 lessons

Who it's for

Is this for you?

Python backend engineers

who have read the MCP spec but never built a server that talks to a real API

AI engineers

who want Claude Desktop to do useful work on their repos, not just answer questions

Platform engineers

who want to expose internal tools to LLM clients through a standard protocol

FAQ

Common questions.

Do I need a paid GitHub plan?
No. A free GitHub account with a personal access token that has repo scope is enough. You can test against your own public repositories.
Which LLM providers are supported?
The workshop ships with a provider abstraction that supports Fireworks, OpenRouter, Gemini, and OpenAI. You only need one API key to follow along.
Do I need Claude Desktop installed?
It helps, because you will wire the server into Claude Desktop, but the FastAPI inspector endpoint lets you test every tool without any MCP client installed.
Is this compatible with Cursor?
Yes. The server speaks stdio MCP, which Cursor and any other compliant MCP client understands. The config block is almost identical to Claude Desktop.

Pricing

Pick the path that fits.

Subscribe to Pro for every paid course, or buy just this one.

Subscribe to Pro

Recommended

Unlock this course and every paid course plus workshop replays. One subscription.

$35$16/mo · Billed annually

You save 54% with regional pricing

Every paid course unlocked
All workshop replays included
New releases as they ship
Cancel anytime

Buy this course

One-time purchase. Lifetime access to every lesson, exercise, and update.

$149$79one-time

You save 47% with regional pricing

Lifetime access to every lesson and future update
Hands-on projects you can show in interviews
Certificate when you finish

Still deciding? Ask Param a question

After this course:

MCP is how LLMs talk to your systems. Build one that does real work.

Enroll

Building MCP servers for GitHub PR review

$79 one-time

Enroll