Coil

소개

AI coding agents often struggle with retaining information between sessions, leading to repetitive debugging and a lack of consistent context. Traditional memory solutions, whether static instruction files or semantic search-based MCP servers, fall short by either requiring manual maintenance, generating false-positive retrievals, lacking a quality signal for useful information, or suffering from cross-project contamination. Coil addresses these critical issues by implementing a structured memory system that leverages typed schemas, allowing only relevant information to be stored, and structured SQL queries against a local SQLite database for precise, context-aware retrieval. This is augmented by a unique usage-based utility scoring mechanism that automatically promotes proven knowledge and filters out noise, ensuring agents recall what truly matters.

주요 기능

• Typed schemas for storing structured memory types (e.g., decision, error, pattern)
• 0 GitHub stars
• Structured SQL queries with filters for precise, context-aware memory retrieval
• Lifecycle hooks for automated memory capture and session context injection
• Automatic project scoping to prevent cross-contamination of memories
• Usage-based utility scoring to surface proven knowledge and decay irrelevant information

사용 사례

• Ensuring agents consistently remember past project decisions, preferences, and established patterns
• Preventing AI coding agents from re-debugging known issues and re-solving past problems
• Automating the capture and retrieval of useful code patterns, errors, and contextual information across development sessions