1# Crush
2
3<p align="center">
4 <a href="https://stuff.charm.sh/crush/charm-crush.png"><img width="450" alt="Charm Crush Logo" src="https://github.com/user-attachments/assets/cf8ca3ce-8b02-43f0-9d0f-5a331488da4b" /></a><br />
5 <a href="https://github.com/charmbracelet/crush/releases"><img src="https://img.shields.io/github/release/charmbracelet/crush" alt="Latest Release"></a>
6 <a href="https://github.com/charmbracelet/crush/actions"><img src="https://github.com/charmbracelet/crush/actions/workflows/build.yml/badge.svg" alt="Build Status"></a>
7</p>
8
9<p align="center">Your new coding bestie, now available in your favourite terminal.<br />Your tools, your code, and your workflows, wired into your LLM of choice.</p>
10<p align="center">终端里的编程新搭档，<br />无缝接入你的工具、代码与工作流，全面兼容主流 LLM 模型。</p>
11
12<p align="center"><img width="800" alt="Crush Demo" src="https://github.com/user-attachments/assets/58280caf-851b-470a-b6f7-d5c4ea8a1968" /></p>
13
14## Features
15
16- **Multi-Model:** choose from a wide range of LLMs or add your own via OpenAI- or Anthropic-compatible APIs
17- **Flexible:** switch LLMs mid-session while preserving context
18- **Session-Based:** maintain multiple work sessions and contexts per project
19- **LSP-Enhanced:** Crush uses LSPs for additional context, just like you do
20- **Extensible:** add capabilities via MCPs (`http`, `stdio`, and `sse`)
21- **Works Everywhere:** first-class support in every terminal on macOS, Linux, Windows (PowerShell and WSL), Android, FreeBSD, OpenBSD, and NetBSD
22- **Industrial Grade:** built on the Charm ecosystem, powering 25k+ applications, from leading open source projects to business-critical infrastructure
23
24## Installation
25
26Use a package manager:
27
28```bash
29# Homebrew
30brew install charmbracelet/tap/crush
31
32# NPM
33npm install -g @charmland/crush
34
35# Arch Linux (btw)
36yay -S crush-bin
37
38# Nix
39nix run github:numtide/nix-ai-tools#crush
40
41# FreeBSD
42pkg install crush
43```
44
45Windows users:
46
47```bash
48# Winget
49winget install charmbracelet.crush
50
51# Scoop
52scoop bucket add charm https://github.com/charmbracelet/scoop-bucket.git
53scoop install crush
54```
55
56<details>
57<summary><strong>Nix (NUR)</strong></summary>
58
59Crush is available via the official Charm [NUR](https://github.com/nix-community/NUR) in `nur.repos.charmbracelet.crush`, which is the most up-to-date way to get Crush in Nix.
60
61You can also try out Crush via the NUR with `nix-shell`:
62
63```bash
64# Add the NUR channel.
65nix-channel --add https://github.com/nix-community/NUR/archive/main.tar.gz nur
66nix-channel --update
67
68# Get Crush in a Nix shell.
69nix-shell -p '(import <nur> { pkgs = import <nixpkgs> {}; }).repos.charmbracelet.crush'
70```
71
72### NixOS & Home Manager Module Usage via NUR
73
74Crush provides NixOS and Home Manager modules via NUR.
75You can use these modules directly in your flake by importing them from NUR. Since it auto detects whether its a home manager or nixos context you can use the import the exact same way :)
76
77```nix
78{
79 inputs = {
80 nixpkgs.url = "github:NixOS/nixpkgs/nixos-unstable";
81 nur.url = "github:nix-community/NUR";
82 };
83
84 outputs = { self, nixpkgs, nur, ... }: {
85 nixosConfigurations.your-hostname = nixpkgs.lib.nixosSystem {
86 system = "x86_64-linux";
87 modules = [
88 nur.modules.nixos.default
89 nur.repos.charmbracelet.modules.crush
90 {
91 programs.crush = {
92 enable = true;
93 settings = {
94 providers = {
95 openai = {
96 id = "openai";
97 name = "OpenAI";
98 base_url = "https://api.openai.com/v1";
99 type = "openai";
100 api_key = "sk-fake123456789abcdef...";
101 models = [
102 {
103 id = "gpt-4";
104 name = "GPT-4";
105 }
106 ];
107 };
108 };
109 lsp = {
110 go = { command = "gopls"; enabled = true; };
111 nix = { command = "nil"; enabled = true; };
112 };
113 options = {
114 context_paths = [ "/etc/nixos/configuration.nix" ];
115 tui = { compact_mode = true; };
116 debug = false;
117 };
118 };
119 };
120 }
121 ];
122 };
123 };
124}
125```
126
127</details>
128
129<details>
130<summary><strong>Debian/Ubuntu</strong></summary>
131
132```bash
133sudo mkdir -p /etc/apt/keyrings
134curl -fsSL https://repo.charm.sh/apt/gpg.key | sudo gpg --dearmor -o /etc/apt/keyrings/charm.gpg
135echo "deb [signed-by=/etc/apt/keyrings/charm.gpg] https://repo.charm.sh/apt/ * *" | sudo tee /etc/apt/sources.list.d/charm.list
136sudo apt update && sudo apt install crush
137```
138
139</details>
140
141<details>
142<summary><strong>Fedora/RHEL</strong></summary>
143
144```bash
145echo '[charm]
146name=Charm
147baseurl=https://repo.charm.sh/yum/
148enabled=1
149gpgcheck=1
150gpgkey=https://repo.charm.sh/yum/gpg.key' | sudo tee /etc/yum.repos.d/charm.repo
151sudo yum install crush
152```
153
154</details>
155
156Or, download it:
157
158- [Packages][releases] are available in Debian and RPM formats
159- [Binaries][releases] are available for Linux, macOS, Windows, FreeBSD, OpenBSD, and NetBSD
160
161[releases]: https://github.com/charmbracelet/crush/releases
162
163Or just install it with Go:
164
165```
166go install github.com/charmbracelet/crush@latest
167```
168
169> [!WARNING]
170> Productivity may increase when using Crush and you may find yourself nerd
171> sniped when first using the application. If the symptoms persist, join the
172> [Discord][discord] and nerd snipe the rest of us.
173
174## Getting Started
175
176The quickest way to get started is to grab an API key for your preferred
177provider such as Anthropic, OpenAI, Groq, OpenRouter, or Vercel AI Gateway and just start
178Crush. You'll be prompted to enter your API key.
179
180That said, you can also set environment variables for preferred providers.
181
182| Environment Variable | Provider |
183| --------------------------- | -------------------------------------------------- |
184| `ANTHROPIC_API_KEY` | Anthropic |
185| `OPENAI_API_KEY` | OpenAI |
186| `VERCEL_API_KEY` | Vercel AI Gateway |
187| `GEMINI_API_KEY` | Google Gemini |
188| `SYNTHETIC_API_KEY` | Synthetic |
189| `ZAI_API_KEY` | Z.ai |
190| `HF_TOKEN` | Hugging Face Inference |
191| `CEREBRAS_API_KEY` | Cerebras |
192| `OPENROUTER_API_KEY` | OpenRouter |
193| `GROQ_API_KEY` | Groq |
194| `VERTEXAI_PROJECT` | Google Cloud VertexAI (Gemini) |
195| `VERTEXAI_LOCATION` | Google Cloud VertexAI (Gemini) |
196| `AWS_ACCESS_KEY_ID` | Amazon Bedrock (Claude) |
197| `AWS_SECRET_ACCESS_KEY` | Amazon Bedrock (Claude) |
198| `AWS_REGION` | Amazon Bedrock (Claude) |
199| `AWS_PROFILE` | Amazon Bedrock (Custom Profile) |
200| `AWS_BEARER_TOKEN_BEDROCK` | Amazon Bedrock |
201| `AZURE_OPENAI_API_ENDPOINT` | Azure OpenAI models |
202| `AZURE_OPENAI_API_KEY` | Azure OpenAI models (optional when using Entra ID) |
203| `AZURE_OPENAI_API_VERSION` | Azure OpenAI models |
204
205### By the Way
206
207Is there a provider you’d like to see in Crush? Is there an existing model that needs an update?
208
209Crush’s default model listing is managed in [Catwalk](https://github.com/charmbracelet/catwalk), a community-supported, open source repository of Crush-compatible models, and you’re welcome to contribute.
210
211<a href="https://github.com/charmbracelet/catwalk"><img width="174" height="174" alt="Catwalk Badge" src="https://github.com/user-attachments/assets/95b49515-fe82-4409-b10d-5beb0873787d" /></a>
212
213## Configuration
214
215Crush runs great with no configuration. That said, if you do need or want to
216customize Crush, configuration can be added either local to the project itself,
217or globally, with the following priority:
218
2191. `.crush.json`
2202. `crush.json`
2213. `$HOME/.config/crush/crush.json`
222
223Configuration itself is stored as a JSON object:
224
225```json
226{
227 "this-setting": { "this": "that" },
228 "that-setting": ["ceci", "cela"]
229}
230```
231
232As an additional note, Crush also stores ephemeral data, such as application state, in one additional location:
233
234```bash
235# Unix
236$HOME/.local/share/crush/crush.json
237
238# Windows
239%LOCALAPPDATA%\crush\crush.json
240```
241
242> [!TIP]
243> You can override the user and data config locations by setting:
244> * `CRUSH_GLOBAL_CONFIG`
245> * `CRUSH_GLOBAL_DATA`
246
247### LSPs
248
249Crush can use LSPs for additional context to help inform its decisions, just
250like you would. LSPs can be added manually like so:
251
252```json
253{
254 "$schema": "https://charm.land/crush.json",
255 "lsp": {
256 "go": {
257 "command": "gopls",
258 "env": {
259 "GOTOOLCHAIN": "go1.24.5"
260 }
261 },
262 "typescript": {
263 "command": "typescript-language-server",
264 "args": ["--stdio"]
265 },
266 "nix": {
267 "command": "nil"
268 }
269 }
270}
271```
272
273### MCPs
274
275Crush also supports Model Context Protocol (MCP) servers through three
276transport types: `stdio` for command-line servers, `http` for HTTP endpoints,
277and `sse` for Server-Sent Events. Environment variable expansion is supported
278using `$(echo $VAR)` syntax.
279
280```json
281{
282 "$schema": "https://charm.land/crush.json",
283 "mcp": {
284 "filesystem": {
285 "type": "stdio",
286 "command": "node",
287 "args": ["/path/to/mcp-server.js"],
288 "timeout": 120,
289 "disabled": false,
290 "disabled_tools": ["some-tool-name"],
291 "env": {
292 "NODE_ENV": "production"
293 }
294 },
295 "github": {
296 "type": "http",
297 "url": "https://api.githubcopilot.com/mcp/",
298 "timeout": 120,
299 "disabled": false,
300 "disabled_tools": ["create_issue", "create_pull_request"],
301 "headers": {
302 "Authorization": "Bearer $GH_PAT"
303 }
304 },
305 "streaming-service": {
306 "type": "sse",
307 "url": "https://example.com/mcp/sse",
308 "timeout": 120,
309 "disabled": false,
310 "headers": {
311 "API-Key": "$(echo $API_KEY)"
312 }
313 }
314 }
315}
316```
317
318### Ignoring Files
319
320Crush respects `.gitignore` files by default, but you can also create a
321`.crushignore` file to specify additional files and directories that Crush
322should ignore. This is useful for excluding files that you want in version
323control but don't want Crush to consider when providing context.
324
325The `.crushignore` file uses the same syntax as `.gitignore` and can be placed
326in the root of your project or in subdirectories.
327
328### Allowing Tools
329
330By default, Crush will ask you for permission before running tool calls. If
331you'd like, you can allow tools to be executed without prompting you for
332permissions. Use this with care.
333
334```json
335{
336 "$schema": "https://charm.land/crush.json",
337 "permissions": {
338 "allowed_tools": [
339 "view",
340 "ls",
341 "grep",
342 "edit",
343 "mcp_context7_get-library-doc"
344 ]
345 }
346}
347```
348
349You can also skip all permission prompts entirely by running Crush with the
350`--yolo` flag. Be very, very careful with this feature.
351
352### Disabling Built-In Tools
353
354If you'd like to prevent Crush from using certain built-in tools entirely, you
355can disable them via the `options.disabled_tools` list. Disabled tools are
356completely hidden from the agent.
357
358```json
359{
360 "$schema": "https://charm.land/crush.json",
361 "options": {
362 "disabled_tools": [
363 "bash",
364 "sourcegraph"
365 ]
366 }
367}
368```
369
370To disable tools from MCP servers, see the [MCP config section](#mcps).
371
372### Agent Skills
373
374Crush supports the [Agent Skills](https://agentskills.io) open standard for
375extending agent capabilities with reusable skill packages. Skills are folders
376containing a `SKILL.md` file with instructions that Crush can discover and
377activate on demand.
378
379Skills are discovered from:
380
381- `~/.config/crush/skills/` on Unix (default, can be overridden with `CRUSH_SKILLS_DIR`)
382- `%LOCALAPPDATA%\crush\skills\` on Windows (default, can be overridden with `CRUSH_SKILLS_DIR`)
383- Additional paths configured via `options.skills_paths`
384
385```jsonc
386{
387 "$schema": "https://charm.land/crush.json",
388 "options": {
389 "skills_paths": [
390 "~/.config/crush/skills", // Windows: "%LOCALAPPDATA%\\crush\\skills",
391 "./project-skills"
392 ]
393 }
394}
395```
396
397You can get started with example skills from [anthropics/skills](https://github.com/anthropics/skills):
398
399```bash
400# Unix
401mkdir -p ~/.config/crush/skills
402cd ~/.config/crush/skills
403git clone https://github.com/anthropics/skills.git _temp
404mv _temp/skills/* . && rm -rf _temp
405```
406
407```powershell
408# Windows (PowerShell)
409mkdir -Force "$env:LOCALAPPDATA\crush\skills"
410cd "$env:LOCALAPPDATA\crush\skills"
411git clone https://github.com/anthropics/skills.git _temp
412mv _temp/skills/* . ; rm -r -force _temp
413```
414
415### Desktop notifications
416
417Crush sends desktop notifications when a tool call requires permission and when
418the agent finishes its turn. They're only sent when the terminal window isn't
419focused _and_ your terminal supports reporting the focus state.
420
421```jsonc
422{
423 "$schema": "https://charm.land/crush.json",
424 "options": {
425 "disable_notifications": false // default
426 }
427}
428```
429
430To disable desktop notifications, set `disable_notifications` to `true` in your
431configuration. On macOS, notifications currently lack icons due to platform
432limitations.
433
434### Initialization
435
436When you initialize a project, Crush analyzes your codebase and creates
437a context file that helps it work more effectively in future sessions.
438By default, this file is named `AGENTS.md`, but you can customize the
439name and location with the `initialize_as` option:
440
441```json
442{
443 "$schema": "https://charm.land/crush.json",
444 "options": {
445 "initialize_as": "AGENTS.md"
446 }
447}
448```
449
450This is useful if you prefer a different naming convention or want to
451place the file in a specific directory (e.g., `CRUSH.md` or
452`docs/LLMs.md`). Crush will fill the file with project-specific context
453like build commands, code patterns, and conventions it discovered during
454initialization.
455
456### Attribution Settings
457
458By default, Crush adds attribution information to Git commits and pull requests
459it creates. You can customize this behavior with the `attribution` option:
460
461```json
462{
463 "$schema": "https://charm.land/crush.json",
464 "options": {
465 "attribution": {
466 "trailer_style": "co-authored-by",
467 "generated_with": true
468 }
469 }
470}
471```
472
473- `trailer_style`: Controls the attribution trailer added to commit messages
474 (default: `assisted-by`)
475 - `assisted-by`: Adds `Assisted-by: [Model Name] via Crush <crush@charm.land>`
476 (includes the model name)
477 - `co-authored-by`: Adds `Co-Authored-By: Crush <crush@charm.land>`
478 - `none`: No attribution trailer
479- `generated_with`: When true (default), adds `💘 Generated with Crush` line to
480 commit messages and PR descriptions
481
482### Custom Providers
483
484Crush supports custom provider configurations for both OpenAI-compatible and
485Anthropic-compatible APIs.
486
487> [!NOTE]
488> Note that we support two "types" for OpenAI. Make sure to choose the right one
489> to ensure the best experience!
490> * `openai` should be used when proxying or routing requests through OpenAI.
491> * `openai-compat` should be used when using non-OpenAI providers that have OpenAI-compatible APIs.
492
493#### OpenAI-Compatible APIs
494
495Here’s an example configuration for Deepseek, which uses an OpenAI-compatible
496API. Don't forget to set `DEEPSEEK_API_KEY` in your environment.
497
498```json
499{
500 "$schema": "https://charm.land/crush.json",
501 "providers": {
502 "deepseek": {
503 "type": "openai-compat",
504 "base_url": "https://api.deepseek.com/v1",
505 "api_key": "$DEEPSEEK_API_KEY",
506 "models": [
507 {
508 "id": "deepseek-chat",
509 "name": "Deepseek V3",
510 "cost_per_1m_in": 0.27,
511 "cost_per_1m_out": 1.1,
512 "cost_per_1m_in_cached": 0.07,
513 "cost_per_1m_out_cached": 1.1,
514 "context_window": 64000,
515 "default_max_tokens": 5000
516 }
517 ]
518 }
519 }
520}
521```
522
523#### Anthropic-Compatible APIs
524
525Custom Anthropic-compatible providers follow this format:
526
527```json
528{
529 "$schema": "https://charm.land/crush.json",
530 "providers": {
531 "custom-anthropic": {
532 "type": "anthropic",
533 "base_url": "https://api.anthropic.com/v1",
534 "api_key": "$ANTHROPIC_API_KEY",
535 "extra_headers": {
536 "anthropic-version": "2023-06-01"
537 },
538 "models": [
539 {
540 "id": "claude-sonnet-4-20250514",
541 "name": "Claude Sonnet 4",
542 "cost_per_1m_in": 3,
543 "cost_per_1m_out": 15,
544 "cost_per_1m_in_cached": 3.75,
545 "cost_per_1m_out_cached": 0.3,
546 "context_window": 200000,
547 "default_max_tokens": 50000,
548 "can_reason": true,
549 "supports_attachments": true
550 }
551 ]
552 }
553 }
554}
555```
556
557### Amazon Bedrock
558
559Crush currently supports running Anthropic models through Bedrock, with caching disabled.
560
561- A Bedrock provider will appear once you have AWS configured, i.e. `aws configure`
562- Crush also expects the `AWS_REGION` or `AWS_DEFAULT_REGION` to be set
563- To use a specific AWS profile set `AWS_PROFILE` in your environment, i.e. `AWS_PROFILE=myprofile crush`
564- Alternatively to `aws configure`, you can also just set `AWS_BEARER_TOKEN_BEDROCK`
565
566### Vertex AI Platform
567
568Vertex AI will appear in the list of available providers when `VERTEXAI_PROJECT` and `VERTEXAI_LOCATION` are set. You will also need to be authenticated:
569
570```bash
571gcloud auth application-default login
572```
573
574To add specific models to the configuration, configure as such:
575
576```json
577{
578 "$schema": "https://charm.land/crush.json",
579 "providers": {
580 "vertexai": {
581 "models": [
582 {
583 "id": "claude-sonnet-4@20250514",
584 "name": "VertexAI Sonnet 4",
585 "cost_per_1m_in": 3,
586 "cost_per_1m_out": 15,
587 "cost_per_1m_in_cached": 3.75,
588 "cost_per_1m_out_cached": 0.3,
589 "context_window": 200000,
590 "default_max_tokens": 50000,
591 "can_reason": true,
592 "supports_attachments": true
593 }
594 ]
595 }
596 }
597}
598```
599
600### Local Models
601
602Local models can also be configured via OpenAI-compatible API. Here are two common examples:
603
604#### Ollama
605
606```json
607{
608 "providers": {
609 "ollama": {
610 "name": "Ollama",
611 "base_url": "http://localhost:11434/v1/",
612 "type": "openai-compat",
613 "models": [
614 {
615 "name": "Qwen 3 30B",
616 "id": "qwen3:30b",
617 "context_window": 256000,
618 "default_max_tokens": 20000
619 }
620 ]
621 }
622 }
623}
624```
625
626#### LM Studio
627
628```json
629{
630 "providers": {
631 "lmstudio": {
632 "name": "LM Studio",
633 "base_url": "http://localhost:1234/v1/",
634 "type": "openai-compat",
635 "models": [
636 {
637 "name": "Qwen 3 30B",
638 "id": "qwen/qwen3-30b-a3b-2507",
639 "context_window": 256000,
640 "default_max_tokens": 20000
641 }
642 ]
643 }
644 }
645}
646```
647
648## Logging
649
650Sometimes you need to look at logs. Luckily, Crush logs all sorts of
651stuff. Logs are stored in `./.crush/logs/crush.log` relative to the project.
652
653The CLI also contains some helper commands to make perusing recent logs easier:
654
655```bash
656# Print the last 1000 lines
657crush logs
658
659# Print the last 500 lines
660crush logs --tail 500
661
662# Follow logs in real time
663crush logs --follow
664```
665
666Want more logging? Run `crush` with the `--debug` flag, or enable it in the
667config:
668
669```json
670{
671 "$schema": "https://charm.land/crush.json",
672 "options": {
673 "debug": true,
674 "debug_lsp": true
675 }
676}
677```
678
679## Provider Auto-Updates
680
681By default, Crush automatically checks for the latest and greatest list of
682providers and models from [Catwalk](https://github.com/charmbracelet/catwalk),
683the open source Crush provider database. This means that when new providers and
684models are available, or when model metadata changes, Crush automatically
685updates your local configuration.
686
687### Disabling automatic provider updates
688
689For those with restricted internet access, or those who prefer to work in
690air-gapped environments, this might not be want you want, and this feature can
691be disabled.
692
693To disable automatic provider updates, set `disable_provider_auto_update` into
694your `crush.json` config:
695
696```json
697{
698 "$schema": "https://charm.land/crush.json",
699 "options": {
700 "disable_provider_auto_update": true
701 }
702}
703```
704
705Or set the `CRUSH_DISABLE_PROVIDER_AUTO_UPDATE` environment variable:
706
707```bash
708export CRUSH_DISABLE_PROVIDER_AUTO_UPDATE=1
709```
710
711### Manually updating providers
712
713Manually updating providers is possible with the `crush update-providers`
714command:
715
716```bash
717# Update providers remotely from Catwalk.
718crush update-providers
719
720# Update providers from a custom Catwalk base URL.
721crush update-providers https://example.com/
722
723# Update providers from a local file.
724crush update-providers /path/to/local-providers.json
725
726# Reset providers to the embedded version, embedded at crush at build time.
727crush update-providers embedded
728
729# For more info:
730crush update-providers --help
731```
732
733## Metrics
734
735Crush records pseudonymous usage metrics (tied to a device-specific hash),
736which maintainers rely on to inform development and support priorities. The
737metrics include solely usage metadata; prompts and responses are NEVER
738collected.
739
740Details on exactly what’s collected are in the source code ([here](https://github.com/charmbracelet/crush/tree/main/internal/event)
741and [here](https://github.com/charmbracelet/crush/blob/main/internal/llm/agent/event.go)).
742
743You can opt out of metrics collection at any time by setting the environment
744variable by setting the following in your environment:
745
746```bash
747export CRUSH_DISABLE_METRICS=1
748```
749
750Or by setting the following in your config:
751
752```json
753{
754 "options": {
755 "disable_metrics": true
756 }
757}
758```
759
760Crush also respects the `DO_NOT_TRACK` convention which can be enabled via
761`export DO_NOT_TRACK=1`.
762
763## Contributing
764
765See the [contributing guide](https://github.com/charmbracelet/crush?tab=contributing-ov-file#contributing).
766
767## Whatcha think?
768
769We’d love to hear your thoughts on this project. Need help? We gotchu. You can find us on:
770
771- [Twitter](https://twitter.com/charmcli)
772- [Slack](https://charm.land/slack)
773- [Discord][discord]
774- [The Fediverse](https://mastodon.social/@charmcli)
775- [Bluesky](https://bsky.app/profile/charm.land)
776
777[discord]: https://charm.land/discord
778
779## License
780
781[FSL-1.1-MIT](https://github.com/charmbracelet/crush/raw/main/LICENSE.md)
782
783---
784
785Part of [Charm](https://charm.land).
786
787<a href="https://charm.land/"><img alt="The Charm logo" width="400" src="https://stuff.charm.sh/charm-banner-next.jpg" /></a>
788
789<!--prettier-ignore-->
790Charm热爱开源 • Charm loves open source