brunch.ai

What It Is

Brunch is an open-source specification format and CLI developed by HASH, aimed at software engineers who write requirements briefs for AI coding agents. It is not a code-generation agent, nor an agent runtime. Instead, it helps teams structure “what they want, which constraints matter, and what counts as a correct result,” so downstream AI coding tools or human developers have to do less guesswork.

Core Capabilities

Based on the official website, Brunch’s core value is requirements clarification: capturing intent in a way that is harder to misinterpret, exposing gaps in the specification before delivery, and making it easier for AI coding agents to follow a brief. Rather than relying on long-form prose like traditional PRDs, user stories, or tickets, it preserves natural language while adding a machine-checkable, human-readable “skeleton.” It can be run via npx brunch, and developer documentation and GitHub source code are available.

AI, Integrations, and Chinese Support

Brunch does not provide large-model capabilities itself, and it is not tied to Claude, OpenAI, or any other model provider. The official site explicitly describes it as agent-agnostic, usable with Claude Code, Cursor, Codex, GitHub Copilot, internal orchestrators, or even well-trained human collaborators. This neutral positioning makes it suitable for teams that mix multiple agents or tools. For Chinese support, the available content does not mention a Chinese interface, Chinese documentation, or localized templates. Chinese-speaking teams can use it, but will need to adapt the language and workflow themselves.

Pricing and Data Privacy

Brunch is dual-licensed under MIT and Apache 2.0. According to the official site, it is free to use for commercial, personal, and any-scale projects, making it highly cost-effective. On privacy, the page does not disclose details about telemetry, cloud processing, data storage, or enterprise compliance. Since it is an open-source CLI and specification format, teams working on sensitive projects should still review the source code, dependencies, and actual runtime behavior before adoption.

Pros, Cons, and Who It’s For

Its strengths are a clear focus, free and open-source licensing, tool neutrality, and the fact that it does not force teams to rewrite their existing requirements process. It can be layered onto a single brief or ticket for trial use. The limitations are that it does not directly improve the underlying model’s capabilities, nor does it guarantee final code quality. If a team still cannot clearly define constraints, Brunch can provide structure, but it cannot replace requirements analysis, testing, or review. It is best suited to engineering teams already using AI programming tools such as Cursor, Claude Code, or Copilot, and who are starting to experience issues such as agent misunderstanding and repeated rework.

Access in China

The official site does not provide information about mainland China access, payments, or mirrors. Since it is free and open source and can be obtained via npm/GitHub, actual usability depends on the local network environment. Alternatives include building an in-house specification template, using traditional PRD/ticket standards, or relying directly on prompt templates built into various AI coding tools.