context-mode keeps raw data out of the LLM context.
Your AI coding agent burns through context. Every gh issue list, every grep, every file read piles
into the conversation and gets re-sent every turn. Fifty turns in, that one command has cost you
750,000 input tokens.
context-mode is an MCP plugin that intercepts large tool output before it touches context. Raw data goes into a local FTS5 store. The agent searches it when it needs to. Same answers. Same work. 30× fewer tokens.
Works on 15 platforms — Claude Code, Cursor, Copilot, Codex, Gemini CLI, JetBrains Copilot, Kiro, and 8 more. 287,000+ developers run it locally today.