Paper Lantern

Gives your AI coding agent access to 2M+ computer science papers through five research tools: explore_approaches surveys technique families with evidence, deep_dive investigates implementation details and failure modes, compare_approaches does side-by-side analysis, and check_feasibility returns a GO/PROTOTYPE/RECONSIDER verdict based on your constraints. Setup is one command via npx that handles OAuth and configures Claude Code, Cursor, Windsurf, Copilot, or other supported agents. The agent automatically reaches for it when making technical decisions about algorithms or architectures, not for syntax or debugging. Essentially turns research literature into executable guidance at decision points.