Captures and reapplies coding insights across sessions to build a persistent knowledge base for agentic workflows.
Session Learning is an intelligent memory system for Claude Code that extracts actionable insights from completed tasks and automatically injects relevant context into new sessions. By categorizing learnings into errors, workflows, tools, architecture, and debugging patterns via structured YAML files, it ensures that hard-won knowledge—such as circular dependency fixes or specific tool quirks—is immediately available when similar challenges arise. This creates a continuous improvement loop that prevents the repetition of past mistakes and streamlines complex development tasks.
Key Features
01Structured YAML storage categorized by errors, workflows, tools, and architecture
02Fail-safe non-blocking design that ensures smooth session transitions
0316 GitHub stars
04Intelligent keyword-based context injection at the start of new sessions
05Automated insight extraction from session transcripts at termination
06Manual management interface via the /amplihack:learnings command
Use Cases
01Preventing recurring configuration or dependency errors across long-running projects
02Automating the documentation of tool-specific gotchas discovered during deep debugging sessions
03Building a project-specific knowledge base of architectural decisions and trade-offs
