Warning
🚧 This is a pre-release under heavy, active development. Things are still in flux but we’re excited to share early progress.
Crush
Crush is a tool for building software with AI.
Installation
Crush has first class support for macOS, Linux, and Windows.
Nightly builds are available while Crush is in development.
- Packages are available in Debian, RPM, APK, and PKG formats
- Binaries are available for Linux, macOS and Windows
You can also just install it with go:
git clone git@github.com:charmbracelet/crush.git
cd crush
go install
Not a developer? Here’s a quick how-to.
Download the latest nightly release for your system. The macOS ARM64 one is most likely what you want.
Next, open a terminal and run the following commands:
cd ~/Downloads
tar -xvzf crush_0.1.0-nightly_Darwin_arm64.tar.gz -C crush
sudo mv ./crush/crush /usr/local/bin/crush
rm -rf ./crush
Then, run Crush by typing crush.
Getting Started
The quickest way to get started to grab an API key for your preferred provider such as Anthropic, OpenAI, or Groq, and just start Crush. You'll be prompted to enter your API key.
That said, you can also set environment variables for preferred providers:
| Environment Variable | Provider |
|---|---|
ANTHROPIC_API_KEY |
Anthropic |
OPENAI_API_KEY |
OpenAI |
GEMINI_API_KEY |
Google Gemini |
VERTEXAI_PROJECT |
Google Cloud VertexAI (Gemini) |
VERTEXAI_LOCATION |
Google Cloud VertexAI (Gemini) |
GROQ_API_KEY |
Groq |
AWS_ACCESS_KEY_ID |
AWS Bedrock (Claude) |
AWS_SECRET_ACCESS_KEY |
AWS Bedrock (Claude) |
AWS_REGION |
AWS Bedrock (Claude) |
AZURE_OPENAI_ENDPOINT |
Azure OpenAI models |
AZURE_OPENAI_API_KEY |
Azure OpenAI models (optional when using Entra ID) |
AZURE_OPENAI_API_VERSION |
Azure OpenAI models |
Configuration
For many use cases, Crush can be run with no config. That said, if you do need config, it can be added either local to the project itself, or globally. Configuration has the following priority:
.crush.jsoncrush.json$HOME/.config/crush/crush.json
LSPs
Crush can use LSPs for additional context to help inform its decisions, just like you would. LSPs can be added manually like so:
{
"$schema": "https://charm.land/crush.json",
"lsp": {
"go": {
"command": "gopls"
},
"typescript": {
"command": "typescript-language-server",
"args": ["--stdio"]
},
"nix": {
"command": "alejandra"
}
}
}
MCPs
Crush supports Model Context Protocol (MCP) servers through three transport types: stdio for command-line servers, http for HTTP endpoints, and sse for Server-Sent Events. Environment variable expansion is supported using $(echo $VAR) syntax.
{
"$schema": "https://charm.land/crush.json",
"mcp": {
"filesystem": {
"type": "stdio",
"command": "node",
"args": ["/path/to/mcp-server.js"],
"env": {
"NODE_ENV": "production"
}
},
"github": {
"type": "http",
"url": "https://example.com/mcp/",
"headers": {
"Authorization": "$(echo Bearer $EXAMPLE_MCP_TOKEN)"
}
},
"streaming-service": {
"type": "sse",
"url": "https://example.com/mcp/sse",
"headers": {
"API-Key": "$(echo $API_KEY)"
}
}
}
}
Logging
Enable debug logging with the -d flag or in config. View logs with crush logs. Logs are stored in .crush/logs/crush.log.
# Run with debug logging
crush -d
# View last 1000 lines
crush logs
# Follow logs in real-time
crush logs -f
# Show last 500 lines
crush logs -t 500
Add to your crush.json config file:
{
"$schema": "https://charm.land/crush.json",
"options": {
"debug": true,
"debug_lsp": true
}
}
Configurable Default Permissions
Crush includes a permission system to control which tools can be executed without prompting. You can configure allowed tools in your crush.json config file:
{
"$schema": "https://charm.land/crush.json",
"permissions": {
"allowed_tools": [
"view",
"ls",
"grep",
"edit:write",
"mcp_context7_get-library-doc"
]
}
}
The allowed_tools array accepts:
- Tool names (e.g.,
"view") - allows all actions for that tool - Tool:action combinations (e.g.,
"edit:write") - allows only specific actions
You can also skip all permission prompts entirely by running Crush with the --yolo flag.
Custom Providers
Crush supports custom provider configurations for both OpenAI-compatible and Anthropic-compatible APIs.
OpenAI-Compatible APIs
Here's an example configuration for Deepseek, which uses an OpenAI-compatible API. Don't forget to set DEEPSEEK_API_KEY in your environment.
{
"$schema": "https://charm.land/crush.json",
"providers": {
"deepseek": {
"type": "openai",
"base_url": "https://api.deepseek.com/v1",
"api_key": "$DEEPSEEK_API_KEY",
"models": [
{
"id": "deepseek-chat",
"name": "Deepseek V3",
"cost_per_1m_in": 0.27,
"cost_per_1m_out": 1.1,
"cost_per_1m_in_cached": 0.07,
"cost_per_1m_out_cached": 1.1,
"context_window": 64000,
"default_max_tokens": 5000
}
]
}
}
}
Anthropic-Compatible APIs
You can also configure custom Anthropic-compatible providers:
{
"$schema": "https://charm.land/crush.json",
"providers": {
"custom-anthropic": {
"type": "anthropic",
"base_url": "https://api.anthropic.com/v1",
"api_key": "$ANTHROPIC_API_KEY",
"extra_headers": {
"anthropic-version": "2023-06-01"
},
"models": [
{
"id": "claude-3-sonnet",
"model": "Claude 3 Sonnet",
"cost_per_1m_in": 3000,
"cost_per_1m_out": 15000,
"cost_per_1m_in_cached": 300,
"cost_per_1m_out_cached": 15000,
"context_window": 200000,
"default_max_tokens": 4096,
"supports_attachments": true
}
]
}
}
}
Whatcha think?
We’d love to hear your thoughts on this project. Feel free to drop us a note!
License
Part of Charm.
Charm热爱开源 • Charm loves open source
