AI Class by Kanary

Product Introduction

1. Definition: AI Class by Kanary is a privacy-centric, local-first analytics tool that performs behavioral analysis on user interactions with AI coding assistants. It technically functions as a log aggregator and pattern recognition engine, processing 30 days of local usage data from GitHub Copilot (Codex) and Claude Code to generate a user profile.
2. Core Value Proposition: It exists to provide AI-powered developers with quantified self-awareness and team benchmarking. By analyzing real interaction logs, it moves beyond generic advice to offer a data-driven personality profile for your AI development workflow, helping you understand your unique collaboration style with code generation models.

Main Features

1. RPG-Style Classification Engine: The core feature analyzes aggregated log metrics to assign one of 16 distinct RPG classes (e.g., Knight, Ninja, Alchemist). This classification is based on a multi-dimensional analysis of user behavior, translating technical data into an intuitive, shareable archetype. It works by applying a rules-based or machine learning model to the six core behavioral metrics.
2. Six Core Behavioral Metrics Dashboard: The tool breaks down your AI assistant usage into six quantitative stats: Volume (total AI-generated tokens/output), Autonomy (average steps AI takes per human instruction), Chat Length (longevity of a single conversation thread), Context Size (amount of code/context provided per prompt), Cache Reuse (frequency of reusing similar prompts/contexts), and Parallelism (number of concurrent AI chats/sessions). These provide a granular, technical view of your working style.
3. Privacy-Preserving Local Analysis: A key technical feature is its local aggregation agent. The software runs on your machine, parsing local IDE or tool logs. It only transmits anonymized, aggregated totals and metric calculations to Kanary's servers. Your raw conversation text, specific prompts, file names, and project details never leave your local environment, addressing major data privacy concerns for developers.

Problems Solved

1. Pain Point: Lack of self-awareness in AI-augmented development. Developers often use tools like Copilot or Claude Code intuitively without understanding their own patterns, leading to inefficient workflows or underutilization of features.
2. Target Audience: The primary user personas are Professional Software Engineers using AI coding assistants daily, Engineering Managers seeking to understand team dynamics and tool adoption, and Tech-Lead Developers interested in optimizing their own and their team's AI pair-programming efficiency.
3. Use Cases: Essential for personal workflow optimization (identifying if you are a "prompt crafter" or a "context manager"), team building and role alignment (forming balanced "parties" with complementary AI classes), and onboarding & training (showing new developers the diverse ways to effectively interact with AI coding tools).

Unique Advantages

1. Differentiation: Unlike generic AI analytics or usage dashboards that focus on cost or raw output volume, AI Class focuses exclusively on behavioral style. It does not rank skill but describes approach, setting it apart from competitive analysis tools. It also differs from manual log review by automating pattern discovery into a simple, gamified framework.
2. Key Innovation: The novel synthesis of local-first privacy, quantitative behavioral analytics, and gamified user archetypes (RPG classes) for professional developers. This approach makes complex log data accessible, engaging, and shareable without compromising sensitive intellectual property or code privacy.

Frequently Asked Questions (FAQ)

1. Is AI Class by Kanary safe for proprietary code? Yes. The tool's core innovation is local log aggregation. It analyzes logs on your machine and sends only high-level, anonymized metric totals (e.g., "context size average: 1500 tokens"). Your actual code, prompts, conversations, and project names are never uploaded or stored on external servers.
2. What do the AI Class RPG classes (Knight, Ninja, etc.) actually mean? Each class represents a distinct behavioral archetype derived from your metric profile. For example, a "Knight" might favor long, deep, single-chat sessions (high Chat Length, low Parallelism), while a "Ninja" might execute many short, precise prompts across multiple tasks (high Parallelism, high Autonomy). They describe your method, not your skill level.
3. How does AI Class by Kanary get my GitHub Copilot and Claude Code data? The tool reads the local usage logs generated and stored on your computer by these AI coding assistants. It does not require API keys or access to your accounts. You must have used these tools locally within the last 30 days for the scan to have data to analyze.
4. Can my AI Class change over time? Absolutely. Your class is a snapshot of your last 30 days of behavior. If you change projects, adopt new prompting strategies, or use different AI tools, your resulting class and metrics will reflect that evolution, making it useful for tracking changes in your development habits.