AI Briefingcontext-mode: MCP Server Reduces 315KB Context to 5.4KB
mksglu/context-mode
About the project
This project solves the problem of massive raw data generated during MCP tool calls rapidly filling the context window. By filtering large outputs such as Playwright snapshots or GitHub issue lists through sandbox tools, it reduces actual context usage by up to 98%.
All events occurring during a session, including file edits, Git operations, and error logs, are stored in SQLite and managed via FTS5 indexing. Even if previous context is lost due to conversation compression, BM25 search accurately restores only the necessary information, enabling agents to resume work from the point of interruption.
It applies the 'Think in Code' paradigm, guiding LLMs to write and execute analysis scripts instead of processing data directly. Instead of reading 50 files, a single script derives the results, reducing token consumption by more than 100x, and supports automatic routing across 17 clients, including Claude Code, Gemini CLI, and VS Code Copilot.
mksglu/context-mode
Context window optimization for AI coding agents. Sandboxes tool output (98% reduction), persists session memory, and enforces routing across 17 platforms via MCP + hooks.
TypeScript
This introduction was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report errors, attribution issues, or removal requests via Contact.