In mid‑October 2026 I announced a capstone project from my MCP course: Eder Ignatowicz’s AI‑powered Code Review Assistant, an open‑source Python FastMCP server that reviews code diffs, flags bugs and security issues, measures complexity, suggests refactorings, and summarizes changes.
- Purpose – Review code diffs for bugs and security issues, analyze quality and complexity, generate refactoring suggestions, and produce concise change summaries.
- Design decision – Use Pydantic for LLM output validation to ensure responses match expected schemas.
- Design decision – Store review guidelines in MCP resources so checklists are reusable across workflows.
- Design decision – Employ workflow patterns (sequential, parallel, conditional) to orchestrate static analysis, LLM review, and optional refactoring steps.
- Design decision – Include a human‑in‑the‑loop UI where developers accept or reject each finding.
- Design decision – Adopt a layered architecture separating the FastMCP API layer from business logic.
- DevOps practices – Ruff linting, pre‑commit hooks, GitHub Actions CI, and Docker containerization enforce code quality and provide reproducible environments.
The open‑source repository is https://lnkd.in/drsDyJi2, and a free seven‑lesson starter guide is available at https://lnkd.in/dGbNfAHY.
- Result – CI passes all linting, type‑checking, and unit tests on each push.
- Reason – Improves testability and maintainability.
- Reason – Balances automation with developer judgment.
- Reason – Supports complex pipelines while keeping each step independent.
- Reason – Centralizes criteria and enables easy updates.
- Reason – Reduces downstream errors caused by malformed LLM outputs.
- Tech stack – Python, FastMCP, Pydantic, Ruff, pre‑commit hooks, GitHub Actions CI, Docker.