Write and modify code in the terminal
Use Crush directly in the CLI for code generation, editing, and debugging
Best for：Terminal-first developers
Coding / Developer
Crush
Crush is a terminal-based agentic coding tool by Charmbracelet that supports switching between different LLM models within the same session while preserving context.
Product type: CLI AgentPricing: UnknownAPIMaster integration: Unknown
OpenRouter token usage rank
#60
Source: OpenRouter
Monthly unique visitors
39,283
Source: SimilarWeb
Overview
Crush is positioned as a CLI tool that lets developers drive AI directly from the terminal to complete coding tasks, primarily interacting with AI through commands to modify code. It supports LSP integration for code context and is compatible with Model Context Protocol extension capabilities. Users can run it across Linux, macOS, Windows, and other platforms, connecting to any OpenAI- or Anthropic-compatible API. Compared with IDE-style AI tools, it is better suited to developers who prefer a pure terminal workflow and do not want to switch between multiple windows. Its core differentiators are a terminal-native experience and flexible model switching.
One-line summary
Crush is a terminal-based agentic coding tool by Charmbracelet that supports switching between different LLM models within the same session while preserving context.
What people use it for
Users enter commands in the terminal to have AI directly edit project code, or switch between different models and continue the same task.
Best for
Developers, engineering teams, and technical leads
How it works
It typically runs in a project environment, IDE, or terminal, completing development tasks by reading context, executing commands, and modifying files.
Quick facts
Product type
CLI Agent
Pricing
Unknown
The current enhanced batch dataset does not maintain real-time pricing for this product. Please refer to the official website or official documentation.
APIMaster integration
Unknown
Pending verification
Data confidence
Medium
Last verified: 2026-08-30
Core features and differentiators
Terminal-native AI coding agent
Runs directly in any terminal emulator and supports a high-performance TUI built with the Charm library
On-demand Docker MCP loading
Dynamically pulls and uses MCP tools through Docker with no configuration required
Non-interactive execution with crush run
Supports calling Crush in non-interactive mode from scripts or Taskfile to complete tasks
Built-in model provider support
Can directly use model services provided by charm.land or a self-hosted provider
Charm ecosystem integration
Built on Charm libraries such as Bubble Tea and Lip Gloss, with support for v2 rendering features
What is Crush best for?
Automate workflows with Taskfile
Integrate AI coding steps into Taskfile via crush run
Best for：DevOps and scripting users
Dynamically invoke Docker MCP tools
Pull specific MCP tools on demand during coding to enhance context
Best for：Developers who need external tools
Switch model providers for testing
Switch between official or self-hosted model providers in Crush to validate outputs
Best for：Developers tuning models
Third-party key setup for Crush (APIMaster integration)
We do not yet have reliable evidence that this product supports third-party keys, proxies, custom providers, or a custom base URL, so we do not show speculative setup steps or config values.
Community discussion
Discussion summary
Users generally understand Crush as a terminal AI coding agent tool launched by the Charm team, emphasizing its high-performance TUI built in Go and deep integration with libraries such as Bubble Tea. Discussion mainly focuses on how to use the tool efficiently in a CLI environment for code editing, command execution, and model interaction, as well as how it fits into existing development workflows. Users also frequently compare it with graphical or JS-based tools such as Cursor and Claude Code, paying attention to resource usage and configuration flexibility.
Most discussed topics
- 1
Crush memory usage and performance
Users often discuss its low RAM usage and fast rendering, making it suitable for running multiple instances on resource-constrained machines or comparing it with Go/Rust tools.
- 2
Integration with Docker, Taskfile, and similar tools
Users explore how to combine it with Docker and Taskfile through non-interactive mode or MCP loading to enable automated workflows and dynamic tool invocation.
- 3
Model providers and LSP/MCP support
Users discuss configuring models such as Anthropic and OpenAI, as well as the setup experience and seamless integration of built-in LSP and MCP servers.
- 4
Using it with nvim or external editors
Users share experiences using Crush in the CLI to replace or complement nvim, along with scenarios where the built-in Vim-like editor is useful.
- 5
Comparison with OpenCode, Cursor, and other AI coding tools
Users compare Crush's terminal aesthetics, session persistence, and file change tracking with the strengths and weaknesses of other agent tools.
FAQ
What type of product is Crush?
We currently classify it under the "Programming / Developer" category, and the page description is based on public information from the official website, OpenRouter, and other sources.
What is Crush suitable for?
The enhanced Crush page prioritizes the core tasks and use cases that have been collected, helping you quickly judge whether it matches your current needs.
Can Crush connect directly to APIMaster now?
There is currently no reliable source confirming that the product supports third-party keys or custom compatible endpoints, so the page status is shown as pending verification.
Similar agents
Claude Code
Also in the Programming / Developer category, useful for side-by-side comparison of different task entry points and product formats.
Cline
Also in the Programming / Developer category, useful for side-by-side comparison of different task entry points and product formats.
Kilo Code
Also in the Programming / Developer category, useful for side-by-side comparison of different task entry points and product formats.
Sources：Official websiteGitHub
Last verified: 2026-08-30 · Report a correction