ToolsCompaniesAbout

Crush

by Charm@charmcliNo ratings yet

A terminal coding agent that works with your code, tools and chosen language model.

Visit github.com

Linked from github.com · checked 1 Nov

Tool type
Source-available terminal coding agent
Made for
Software developers
Runs on
Terminal on macOS, Linux, Windows, Android, and BSD variants
Last checked
1 Nov 2026
4 product sources

From @charmcli, about Crush

Recent important posts · launches

What it does

Crush is a terminal-based AI coding agent for software developers. It works in a project directory and can use local development tools, code, and workflows as context. Users can choose among supported language models or connect their own through OpenAI- or Anthropic-compatible APIs, then switch models during a session while keeping context. Crush supports multiple project sessions, uses language servers for additional context, and can extend its available tools through Model Context Protocol (MCP) servers.

Crush runs locally in terminals on macOS, Linux, Windows, Android and BSD systems. Install a package or binary, open a project, and connect a supported model provider or a compatible local endpoint. Model access can require separate accounts, subscriptions or usage payments. The current client uses the Functional Source License, with restrictions on competing services and an MIT conversion after two years for each release. Tool calls normally require permission, although some read-only commands do not. Review repository configuration and requested commands because the agent can edit files and run development tools.

What it can do

  • Coding agent with access to project code and local development tools
  • Choose among multiple language models or configure a compatible provider
  • Switch models during a session while preserving context
  • Maintain multiple work sessions and project contexts
  • Use language-server protocols for additional code context
  • Extend tools through MCP servers using stdio, HTTP, or SSE

Works with

  • OpenAI-compatible APIs
  • Anthropic-compatible APIs
  • Model Context Protocol (MCP) servers
  • Language Server Protocol (LSP)

Before you choose

  • Current releases use FSL-1.1-MIT, with competing-use restrictions; do not assume unrestricted MIT terms apply immediately.
  • Crush can execute tools against project files; tool execution prompts are enabled by default, but permissions can be changed and --yolo bypasses prompts.
  • Treat crushrc and crush.json as trusted code: crushrc runs in a full shell, and the repository warns against launching Crush in a directory with unreviewed configuration.

Listed under

Get started

  1. Install Crush using a listed package manager, download a binary, or install with Go.
  2. Launch Crush in the project directory you want it to work with.
  3. Choose a model and authenticate with its provider, or select another provider and supply its API key.
Install Crush and connect a model

Plans and prices

The current client license permits internal use but restricts competing services. Hyper is an optional paid provider; other API providers or local compute have their own costs.

We couldn’t find published prices for Crush. See their pricing page ↗

Sources for this overview

Product pages, documentation, and pricing checked 1 Nov 2026. Each source supports part of the overview; availability and prices can change.

Ratings and reviews

No ratings for Crush yet.

Be the first to rate it

What’s changed

Free account

See what’s changed at Crush

Launches, price moves and pivots, each with a date and a source. Free with an account.

Signals

Free account

See Crush’s signals

Launch traction, hiring and release activity, tracked over time. Free with an account.

Data and security

Free account

See how Crush handles your data

Training on your files, compliance, hosting and the models behind it. Free with an account.

Compare with similar tools

Free account

Compare Crush side by side

Prices, platforms and features against similar tools, with sources. Free with an account.

Where this comes from

How we research
Company website
github.com: product, pricing and policy pages
YC directory mirror
Batch and company name
Social accounts
Accounts the company links, and its posts on X
Human review
Checked by a person
knowledge_base
manual_review

Last checked 1 Nov 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.