Crush Tutorial: Multi-Model Terminal Coding Agent with Strong Extensibility

June 15, 2026 ยท View on GitHub

Learn how to use charmbracelet/crush for terminal-native coding workflows with flexible model providers, LSP/MCP integrations, and production-grade controls.

GitHub Repo License Docs

Why This Track Matters

Crush is one of the fastest-growing terminal coding agents and is designed around practical developer loops: model flexibility, project sessions, local config control, and advanced integrations.

This track focuses on:

  • installing and configuring Crush for immediate coding work
  • controlling providers, permissions, and tool behavior safely
  • integrating LSP and MCP for richer code intelligence
  • standardizing Crush usage across team environments

Current Snapshot (auto-updated)

  • repository: charmbracelet/crush
  • stars: about 25.3k
  • GitHub release reference: v0.77.0 (checked 2026-06-15; release metadata on GitHub)

Mental Model

flowchart LR
    A[Developer Prompt] --> B[Crush Session]
    B --> C[Provider and Model Router]
    C --> D[Tool Calls LSP/MCP/Built-in]
    D --> E[File and Command Effects]
    E --> F[Review, Iterate, Commit]

Chapter Guide

ChapterKey QuestionOutcome
01 - Getting StartedHow do I install and run Crush fast?Working baseline
02 - Architecture and Session ModelHow does Crush manage project and session state?Runtime mental model
03 - Providers and Model ConfigurationHow do I configure providers and models robustly?Stable provider strategy
04 - Permissions and Tool ControlsHow do I enforce safe tool execution boundaries?Safer execution model
05 - LSP and MCP IntegrationHow do I plug in language servers and MCP tools?Extended capabilities
06 - Skills, Commands, and Workflow CustomizationHow do I tailor Crush to team and project conventions?Reusable workflow system
07 - Logs, Debugging, and OperationsHow do I diagnose and maintain Crush at scale?Operational confidence
08 - Production Governance and RolloutHow do teams adopt Crush safely in production repos?Governance runbook

What You Will Learn

  • how to configure Crush for consistent coding-agent performance
  • how to manage provider routing and custom model definitions
  • how to extend Crush with LSP, MCP, skills, and custom commands
  • how to operate Crush with robust production controls

Source References


Start with Chapter 1: Getting Started.

Full Chapter Map

  1. Chapter 1: Getting Started
  2. Chapter 2: Architecture and Session Model
  3. Chapter 3: Providers and Model Configuration
  4. Chapter 4: Permissions and Tool Controls
  5. Chapter 5: LSP and MCP Integration
  6. Chapter 6: Skills, Commands, and Workflow Customization
  7. Chapter 7: Logs, Debugging, and Operations
  8. Chapter 8: Production Governance and Rollout

Generated by AI Codebase Knowledge Builder