# CrabTalk > An tiny 8 MB agent daemon that manages sessions, dispatches commands, and streams events — install, attach, and build your own AI stack with no telemetry or vendor lock-in. CrabTalk is a lightweight 8 MB agent daemon that installs in seconds and lets you attach an AI assistant directly from your terminal. It manages sessions, dispatches commands, and streams events, while keeping everything else modular and outside the core. CrabTalk is built for developers who want full control over their AI stack without bloat, telemetry, or vendor lock-in. - **Minimal footprint** — *The daemon weighs just 8 MB, handling sessions, dispatch, and events without bundling dozens of tools you didn't ask for.* - **Install & attach workflow** — *Run `curl -sSL https://crabtalk.ai/install | sh`, then `crabtalk daemon install` and `crabtalk attach` to get a live AI chat session in your terminal.* - **Skills & MCP support** — *Extend the daemon with Skills and MCP servers to add only the capabilities you need.* - **Light built-in memory** — *Maintains context across sessions so the agent remembers relevant information between conversations.* - **Cargo-style commands** — *Familiar command structure lets you swap, remove, or write your own commands with ease.* - **No telemetry** — *Nothing phones home; your data stays private and local.* - **Composable stack** — *Connect any binary on your PATH — search, Telegram, gateways, MCP servers — to build a fully custom agent stack.* - **No single point of failure** — *Modular architecture means no one component can take down your entire workflow.* - **Cross-platform** — *Supports macOS, Linux, and Windows environments.* ## Features - 8 MB agent daemon - Session management - Command dispatch - Event streaming - Skills & MCP server support - Light built-in memory - Cargo-style commands - No telemetry - Composable stack - Cross-platform support ## Integrations MCP servers, Telegram, Any binary on PATH ## Platforms CLI, MACOS, LINUX, WINDOWS ## Pricing Free ## Links - Website: https://crabtalk.ai - Documentation: https://www.crabtalk.ai/docs - Repository: https://github.com/crabtalk/crabtalk - EveryDev.ai: https://www.everydev.ai/tools/crabtalk