Mnemonic is a sophisticated memory server designed to give AI coding assistants, such as GitHub Copilot, a persistent 'brain.' It integrates with Copilot via the MCP protocol, allowing the AI to store, retrieve, and analyze project-specific information and past conversations. Utilizing a token-efficient, three-layer progressive disclosure search system (index → timeline → details) and SQLite with FTS5, Mnemonic ensures fast, scalable access to relevant context. This enables the AI to extract decisions, conventions, and bug patterns, populate baseline architectural memories, and even generate documentation like ADRs and onboarding guides, transforming the editor into a more intelligent and informed development environment.
主な機能
01Progressive disclosure search system for efficient token usage
02SQLite + FTS5 for fast, scalable full-text search of memories
03Auto-analysis of conversations to extract decisions and patterns
04Project scanner to populate baseline architectural memories from codebase
05Automated documentation generation (ADRs, onboarding guides, changelogs)
060 GitHub stars
ユースケース
01Enhancing AI coding assistants like GitHub Copilot with persistent, context-aware memory.
02Facilitating efficient knowledge retrieval and context for developer queries within the editor.
03Automatically generating architectural decision records, onboarding guides, and changelogs from accumulated knowledge.
