Crush
A terminal coding agent from the team that makes terminal interfaces look good.
Crush comes from Charm, whose TUI libraries sit under a large share of modern terminal tools. It is a model-agnostic coding agent, and its distinguishing quality is interface: the session is markedly more readable than the category norm. Whether that matters depends on how many hours a day you spend watching an agent work.
Crush gets picked for a reason people are slightly embarrassed to say out loud: how the output looks. That sounds trivial until you have spent four hours reading agent sessions, at which point a badly formatted transcript is genuinely worse to work with. It is model-agnostic like the others in its family, so the usual configuration tax applies. Pick it if you watch agent output for hours rather than minutes; the ergonomics compound at that duration and barely register below it.
Our take
Pick it if you read agent output all day and the readability of that output is a real cost you feel. That is a narrower reason than it sounds, but it is a legitimate one — watching a badly formatted agent session for hours is genuinely worse. On raw capability it sits with the other model-agnostic agents rather than ahead of them, so choose on interface, not on power.
Last checked 2026-09-02
Best for
- People who spend hours a day reading agent sessions
- Anyone already invested in the Charm ecosystem
- Model-agnostic setups that also care about presentation
Limitations
- Capability is comparable to peers, not ahead of them
- Smaller ecosystem than the most-adopted open-source agents
- You supply and pay for the model
When to pick something else
- Teams choosing on capability benchmarks alone
- Anyone who wants the largest plugin ecosystem
- Non-terminal workflows
Key features
- Model-agnostic, bring your own provider
- Unusually readable terminal interface
- Session-based workflow
- Written in Go, single binary
Crush alternatives
If Crush is not the right fit, these solve similar problems.
- opencodeOpen-source terminal coding agent that runs on whichever model you point it at.Official: opencode.aiFree
- Claude CodeTerminal-based coding agent that runs commands and edits files directly.Official: claude.com/product/claude-codefrom $20/mo
- gooseOpen-source agent that installs, executes and tests — not just suggests code.Official: goose-docs.aiFree
- ChatGPTThe most widely used assistant, with image generation and a large plugin ecosystem.Official: chatgpt.comfrom $20/mo
- ClaudeAssistant known for long-context reasoning and writing that needs less editing.Official: claude.aifrom $20/mo
- GeminiGoogle's assistant, tightly integrated with Search, Gmail and Workspace.Official: gemini.google.comfrom $20/mo