# betterCTX > betterCTX (bctx) is a token compression engine for AI coding tools. It runs locally as a CLI, shell hook, and MCP server to cut LLM token consumption by up to 99% across Cursor, Claude Code, Copilot, Windsurf, and any MCP-compatible agent. bctx processes code and conversation context entirely on-device. It exposes 41 skills (Wave 1 + Wave 2 + Wave 3) across reading, code intelligence, memory, session, execution, and analysis, and supports 17 languages via tree-sitter. ## Product - [Home](/home): Overview, quickstart, and live token savings. - [How it works](/how-it-works): Compression engine, lens modes, and the shell hook. - [Benchmark](/benchmark): Measured token savings across real coding sessions. - [Compatibility](/compatibility): 24 native agent integrations and any MCP client. - [Extensions](/extensions): Editor and shell extensions. - [Pricing](/pricing): Free local tier, Beacon+ ($9/mo), Studio ($29/mo), and Let's Talk. - [Changelog](/changelog): Release notes for the bctx CLI and MCP server. - [Compare](/compare): How bctx compares to other context tools. ## Core concepts - [Lens modes](/lens-modes): Compression strategies per task type. - [Shell hook](/shell-hook): Transparent compression at the terminal layer. - [Vault](/vault): Local-only encrypted context store. - [Token savings](/token-savings): Methodology behind the savings numbers. - [MCP server](/mcp-server): Local MCP transport for any compatible agent. - [Multi-agent](/multi-agent): Coordinating multiple agents through bctx. - [Protocols](/protocols): CCP, CEP, and TDD symbol shorthands. - [Context engineering](/context-engineering): The discipline behind bctx. ## Documentation - [Getting started](/docs/getting-started): Install bctx and run your first session. - [Quick reference](/docs/quick-reference): Commands and flags at a glance. - [Configuration](/docs/configuration): Project and global config. - [CLI reference](/docs/cli): Full CLI surface. - [Skills (MCP)](/docs/tools): Overview of the 41 skills. - [Reading skills](/docs/tools/reading): File and repo reading skills. - [Code intelligence](/docs/tools/intelligence): Symbol, dependency, and structural skills. - [Memory skills](/docs/tools/memory): Long-term and project memory skills. - [Session skills](/docs/tools/session): Live conversation skills. - [Execution skills](/docs/tools/execution): Run, test, and verify skills. - [Analysis skills](/docs/tools/analysis): Diff, lint, and audit skills. - [Code intelligence engine](/docs/tree-sitter): Tree-sitter parsing details. - [Analytics](/docs/analytics): Local-only usage insights. - [Token guide](/docs/token-guide): How tokens are counted and saved. ## Guides - [Troubleshooting](/docs/troubleshooting): Fix common issues. - [Writing recipes](/docs/writing-recipes): Patterns for high-leverage prompts. - [CI/CD integration](/docs/ci-cd): Run bctx in pipelines. - [Team setup](/docs/team-setup): Share config across a team. - [Switching to bctx](/docs/migrating): Migrate from other context tools. ## Reference - [MCP API schemas](/docs/mcp-api): Full MCP request/response schemas. - [Security model](/docs/security-model): 100% local processing guarantees. - [Supported languages](/docs/supported-languages): The 17 languages bctx parses. ## Blog - [Blog index](/blog): Articles on context engineering and token economics. ## Optional - [Privacy](/privacy): Privacy policy. - [Contact](/contact): Get in touch.