context-mode: Context window optimization tool for AI coding agents

context-mode is an MCP server that targets the context window problem in AI coding agents. It keeps raw data out of context by sandboxing tools—for example, Playwright snapshots shrink from 56 KB to a tiny size, and 315 KB can be reduced to 5.4 KB, a 98% reduction. For session continuity, every file edit, git operation, task, error and user decision is recorded in SQLite; when conversations are compacted, events are indexed into FTS5, and BM25 search retrieves only relevant content so the model can resume where it left off. It also forces agents to compute with code rather than data—for instance, counting file lines with a script instead of 47 read operations, saving 100x context. The tool supports 17 platforms including Claude Code, Gemini CLI, VS Code Copilot, JetBrains Copilot and GitHub Copilot CLI. Claude Code developers can install via the plugin marketplace and automatically get all hooks and 11 MCP tools; Gemini CLI users need only one config file to enable hooks. This reduces context exhaustion and memory loss in long conversations, making it especially suited to dev teams handling large numbers of files or tasks lasting hours.