Yash Gajjar

Background

Yash Gajjar is an AI/ML engineer focused on developer tooling, retrieval systems, and context optimization for large language models. He created CostWise-MCP to solve the dominant cost in long AI coding sessions — Anthropic prompt-cache write/read charges — by keeping tokens out of the context window entirely.

CostWise-MCP is the first MCP server to introduce stash_context and recall tools that park large content on disk instead of the conversation window, and a session_brief tool that lets agents catch up without re-reading history. The project has grown to 11 MCP tools across retrieval, maintenance, and context-control categories.

Yash's work spans machine learning, systems programming in Go, and open-source infrastructure for AI coding agents. He maintains the project on GitHub and publishes research on context compression and repository intelligence.

Expertise