Awesome-Context-Engineering

Why we included this project

If you build LLM applications or agents, you've probably hit the same wall: prompts that work in demos but drift in production. This curated survey maps the move from static prompting to context engineering, and it goes beyond the usual long-context, RAG, and memory topics into the newer agent-era concerns like runtimes, protocols (MCP, A2A), coding-agent memory, and observability. Each section collects papers, frameworks, and implementation guides, so you can move from understanding a concept to finding concrete tooling to try. The organization is backed by a published arXiv survey, which gives it academic grounding rather than the feel of a random link dump. Teams designing context-heavy pipelines or evaluating agent stacks will find it a practical map for deciding what to read and what to prototype next.