microsoft

AI-Engineering-Coach: Analyze AI Coding Assistant Usage

AI Engineer Coach analyzes local session logs from AI coding assistants and presents practice, activity, and output insights in a dashboard. It suits developers who want to improve agentic workflows while keeping session data on their machine.

TypeScriptMITDeveloper ToolsAI AgentsDashboard
View on GitHub Compare
Stars
4.3k
Forks
597
Stars, last 7 days
+8

Our take

Health score

53/100

How it is scored
Activity
43
Community
70
Issues
53
Pull requests
46

+8 stars in the last 7 days

Overview

AI Engineer Coach is a dashboard for reviewing how you work with AI coding assistants. It reads local session logs across supported harnesses and surfaces patterns in activity, generated code, prompts, and workspace context, helping developers identify habits to improve without sending session data off-device.

It is most useful for individuals who want to inspect and improve their AI-assisted engineering practice. The project is a VS Code extension and also offers a canvas in the GitHub Copilot app, with some AI-assisted features unavailable in canvas mode.

Key Features

  • Summarizes practice scores, activity trends, and session timelines.
  • Detects anti-patterns with 45 editable rules covering prompt quality, session hygiene, code review, tool use, and context management.
  • Breaks down AI-generated code volume by language, workspace, model, and harness.
  • Finds repeated prompts that may be turned into reusable skills.
  • Audits instruction files and maps workspace context, with optional AI review.
  • Provides activity patterns, learning exercises, achievements, and shareable summaries.
  • Analyzes local logs read-only, with no proprietary telemetry or session data sent by the extension.

Use Cases

  • A developer wants to understand whether their AI-assisted coding habits are improving over time.
  • A team member wants to spot recurring workflow issues, such as weak prompts or poor session hygiene, and try concrete changes.
  • A developer working across multiple AI coding harnesses wants to compare activity and generated output in one dashboard.
  • A user wants to audit project instructions and workspace context before relying more heavily on an AI coding assistant.

Getting Started

The extension is not published to a marketplace or Releases page. Build and install the VSIX locally:

git clone https://github.com/microsoft/ai-engineering-coach.git
cd ai-engineering-coach
npm ci
npm run package

Install the generated VSIX in VS Code, then run AI Engineer Coach: Open Dashboard from the command palette. See the official repository README for Dev Container instructions and Copilot app canvas setup.

Alternatives

  • sniffly: Sniffly focuses on Claude Code logs, usage patterns, errors, and message review, while AI-Engineering-Coach analyzes local assistant sessions for practice and output insights.
ProjectLanguageLicenseStarsStatus
AI-Engineering-CoachTypeScriptMIT4.3kNot checked yet
snifflyPythonMIT1.3kInactive

Considerations

  • Requires VS Code 1.115 or later for the extension; building locally requires Node.js and npm. The alternative Dev Container path requires the Dev Containers extension and Docker or Podman.
  • Installation requires building a VSIX yourself because the extension is not published to a marketplace or Releases page.
  • Some features use the VS Code built-in Copilot language model API only when explicitly invoked. Skill Finder, Learning Center, Level Up, and Context Health AI review are unavailable in canvas mode.
  • GitHub issue credit figures are rough estimates based on local usage and inferred issue links, not billing figures.
  • The README describes the project as an open-source community effort by Microsoft employees, not an official Microsoft product or support offering.

Found this useful?

Share it with someone who would like AI-Engineering-Coach.

Comparisons

sniffly vs AI-Engineering-Coach