AI Code Memory Index

A queryable project memory for AI-heavy developers who forget what they built a week ago

Problem

Developers report that when letting AI write large projects, after just a weekend they completely forget where a module from last week is, how it works, and how data flows. When a colleague asks whether a certain logic is included, they can't recall it at all and have to ask AI to re-analyze the code, causing real anxiety. This is a new kind of loss of control.

Solution

A lightweight, local-first code memory tool: it automatically scans Git commits and file changes, generates structured module summaries (what it does, key files, data flow, dependencies) after each AI-assisted coding session, and stores them in a searchable index. It supports fast recall by feature, file, or time, plus natural-language Q&A like 'where is this feature implemented?', so developers can grasp the project without reading thousands of lines of freshly AI-generated code.

Users

Mid-to-senior indie developers and tech leads on small teams who lean heavily on AI coding assistants (Claude Code, Cursor, Copilot, etc.) and see their projects growing in modules and mental overhead.

MVP

Watch specified Git repos for commits and file changes; call a local or cloud LLM to generate module-level summaries for new changes; store summaries and code locations in SQLite; provide a CLI or minimal web UI for keyword search and natural-language Q&A; output 'memory cards' from change history.

Revenue

Free local version limits indexed projects or history depth; Pro subscription unlocks unlimited projects, natural-language Q&A, and shared team memory; optional API usage-based pricing.

Why now

AI coding assistants produce code far faster than humans can remember, and existing code search or IDE plugins only solve 'finding code', not 'remembering intent'. Context windows remain limited, so external memory is needed. Indie developers can build a working version in weeks using existing LLM APIs.