cando.sh

What It Is

CanDo is a local coding Agent for developers. Its website describes it as “not another chatbot,” but a tool that can generate real code, execute real commands, and deliver working features. It provides installation scripts for Linux, macOS, and Windows, and also offers a Beta version for testers.

Core Capabilities

Based on the scraped text, CanDo’s core focus is autonomous development: it can write code, build features, and run code without the user constantly supervising it. Its model capabilities are not tied to a single provider. It supports GLM models via Z.AI, and can also access Claude, GPT-4, and 100+ other models through OpenRouter. It also emphasizes the ability to remember codebase context and resume previous coding sessions even after the computer has been turned off, which is valuable for multi-round iterative development.

Typical Use Cases and Output Quality

The official examples include generating a Doom-style game, a Pacman game, 3D/galaxy animations, moon phase animations, and a sculpting tool within minutes. This suggests CanDo is better suited for rapid prototyping, demos, small tools, and feature iteration. Its strength is a clear goal: delivering runnable code rather than just offering suggestions. However, the text does not disclose its testing process, code quality evaluation, security sandboxing, or failure-handling mechanisms, so complex production-grade projects still require human review.

Pricing, Privacy, and Integrations

The website content does not provide information on pricing, free quotas, trial policies, or payment methods. In terms of integrations, it explicitly mentions Z.AI, OpenRouter, GitHub, and Discord; it does not state whether an official API is available. Data privacy details are also missing, such as whether code is uploaded, how context is stored, and whether local-only processing is supported. This is a major gap for enterprise evaluation.

Pros, Cons, and Best-Fit Users

Its strengths include cross-platform support, simple installation, flexible model selection, Agent-style execution, and contextual memory. Its drawbacks are the lack of disclosure around key details, the possibility of bugs in the Beta version, and the security and misoperation risks that come with autonomous command execution. It is suitable for indie developers, hackathon teams, early adopters of AI coding tools, and users who need to build prototypes quickly.

Access from China

Access from China is currently unclear. Because it relies on overseas services such as GitHub installation scripts, the Discord community, and OpenRouter/Claude/GPT-4, users in mainland China may face uncertainty around network connectivity, model API access, and payments. Comparable alternatives include Cursor, GitHub Copilot, Claude Code, Aider, and OpenHands.