@Xudong07452910: 强烈推荐的开源项目:Trellis —— 让你的 AI Coding Agent 拥有团队级工程规范的开源 Harness Trellis 是我一直在使用的 AI 编码工程框架,被称为“The best agent harness”。 它…
摘要
Trellis 是一个开源项目,为 AI 编码代理提供团队级工程规范和项目记忆框架,支持多种平台,通过将规范和任务上下文持久化到代码仓库提升 Agent 的开发效率和质量。
查看缓存全文
缓存时间: 2026/06/12 02:50
强烈推荐的开源项目:Trellis —— 让你的 AI Coding Agent 拥有团队级工程规范的开源 Harness
Trellis 是我一直在使用的 AI 编码工程框架,被称为“The best agent harness”。
它通过将项目规范、任务上下文和记忆持久化到代码仓库,让 AI Agent(如 Claude Code、Codex、Hermes Agent 等)不再每次从零开始,而是拥有团队级的工程标准和跨平台一致性,大幅提升开发效率和代码质量。
具体来说:
很多人觉得 AI 编码 Agent 越写越不靠谱,第一反应是换模型、堆提示词、加更长的上下文。但问题很多时候不在模型,而在它每次进项目时都是“空着脑子”进场:
项目结构不知道,命名规范不知道,技术选型不知道,上次做到哪也不知道。
Trellis 做的事,就是给 AI 编码 Agent 装上一套项目级工程记忆。
它会在你的代码库里建立一个 .trellis/ 目录,把项目规范、任务 PRD、实现上下文、审查记录、会话日志都沉淀下来,并跟代码一起进入版本管理。
也就是说,你不是每次重新教 AI 怎么写这个项目,而是让它每次开工前先读项目自己的“工程手册”。
它最有价值的地方不是单纯塞上下文,而是把 AI 编码流程变成一套标准工作流:
- Plan:先把需求问清楚
- Implement:按 PRD 和项目规范写代码
- Verify:对照 lint、type check、test 和 spec 自查
- Finish:把这次踩过的坑和新规则写回项目记忆
这就很像把一个“聪明但每天失忆的临时工”,慢慢训练成一个“知道项目规矩、能接着上次继续干活的团队成员”。
更关键的是,Trellis 不绑死工具。它支持 Claude Code、Codex、Cursor、OpenCode、Gemini CLI、GitHub Copilot 等 14 个 AI 编码平台。团队里一个人沉淀的规范,全队的 AI 都能继承。
核心特性:
- 自动注入规范:.trellis/spec/ 中编写一次,AI 每次会话自动加载
- 任务中心工作流:.trellis/tasks/ 存放 PRD、实现上下文、审查记录
- 项目记忆:.trellis/workspace/ 记录上一次会话状态
- 团队共享标准:规范随代码版本控制,团队共用
目前已获 9.7k+ stars ,有中英文文档,社区活跃。特别适合个人开发者、团队工程师和重度使用 AI 编码工具的用户。
也推荐一下项目的负责人之一:卡牌大师崔斯特 @cuisitekp。
https://github.com/mindfold-ai/Trellis…
#AIAgent #ClaudeCode #AgentHarness #AI工具 #开源项目 #Codex
mindfold-ai/Trellis
Source: https://github.com/mindfold-ai/Trellis
An out-of-the-box engineering framework for AI coding.
AI writes code fast, but every session it starts from scratch — no memory of your project, your conventions, or your team's requirements. Trellis persists specs, tasks, and memory into your repo, so any coding agent works to your engineering standards.
简体中文 • Docs • Quick Start • Supported Platforms • Use Cases
Why Trellis?
| Capability | What it changes |
|---|---|
| Auto-injected specs | Write conventions once in .trellis/spec/, then let Trellis inject the relevant context into each session instead of repeating yourself. |
| Task-centered workflow | Keep PRDs, implementation context, review context, and task status in .trellis/tasks/ so AI work stays structured. |
| Project memory | Journals in .trellis/workspace/ preserve what happened last time, so each new session starts with real context. |
| Team-shared standards | Specs live in the repo, so one person’s hard-won workflow or rule can benefit the whole team. |
| Multi-platform setup | Bring the same Trellis structure to 14 AI coding platforms instead of rebuilding your workflow per tool. |
Prerequisites:
- Node.js >= 18
- Python >= 3.9
Quick Start
# 1. Install Trellis
npm install -g @mindfoldhq/trellis@latest
# 2. Initialize in your repo
trellis init -u your-name
# 3. Or initialize with the platforms you actually use
trellis init --cursor --opencode --codex -u your-name
See the Quick Start and Supported Platforms guides for setup details.
How to Use
The workflow is simple:
- Describe what you want in natural language.
- Brainstorm with the AI one question at a time until the PRD is clear, then implementation begins.
- Let it run — the AI calls Trellis Implement and auto-checks the result against specs, lint, type-check, and tests.
- Type
/trellis:finish-workwhen the work is done or the session context fills up. Trellis archives the task and updates journals.
How It Works
Trellis runs a 4-phase loop with auto-invoked skills and sub-agents:
- Plan —
trellis-brainstormwalks through requirements one question at a time and writesprd.md. Research-heavy items go to atrellis-researchsub-agent. The result is curated specs + research files referenced fromimplement.jsonl/check.jsonl. - Implement — a
trellis-implementsub-agent writes code from the PRD with the curated context auto-injected, no git commit. - Verify — a
trellis-checksub-agent reviews the diff against specs and runs lint, type-check, and tests, self-fixing where it can. - Finish — a final check runs, then
trellis-update-specpromotes new learnings back into.trellis/spec/so the next session starts smarter.
Resources
| Need | Link |
|---|---|
| Install Trellis in a repo | Quick Start |
| Understand platform differences | Supported Platforms |
| See the workflow in practice | Real-World Scenarios |
| Start from spec templates | Spec Templates |
| Track releases | Changelog |
FAQ
How is Trellis different from CLAUDE.md, AGENTS.md, or .cursorrules?
Those files are useful entry points, but they tend to become monolithic. Trellis adds scoped specs, task PRDs, workflow gates, workspace memory, and platform-aware generated files around them.
Is Trellis only for Claude Code?
No. Trellis is a project layer that works across multiple coding agents and IDEs.
Is Trellis for solo developers or teams?
Both. Solo developers use it for memory and repeatable workflow. Teams get the larger benefit: shared standards, task boundaries, reviewable context, and platform portability.
Do I have to write every spec file manually?
No. Many teams start by letting AI draft specs from existing code and then tighten the important parts by hand. Trellis works best when you keep the high-signal rules explicit and versioned.
Can teams use this without constant conflicts?
Yes. Personal workspace journals stay separate per developer, while shared specs and tasks stay in the repo where they can be reviewed and improved like any other project artifact.
Star History
Community & Resources
Official Repository • AGPL-3.0 License • Built by Mindfold
相似文章
@cuisitekp: 长期用 Claude Code 或 Codex 写项目的人,真该把 Trellis 装上试试。 说它是现在最接近"让 AI 记住你项目"的方案,不算夸张。 很多人写着写着觉得 AI 越来越不靠谱,第一反应是去换更强的模型、或者把提示词堆得…
Trellis is an open-source engineering framework that persists project context, specs, tasks, and memory into a repository, enabling AI coding agents to remember conventions and progress across sessions. It integrates with 14 AI coding platforms and aims to solve the problem of AI forgetting project context, improving development workflow for teams and individuals.
@wsl8297: 在 GitHub 上发现一个专门给团队用的 AI 编程管理工具:Trellis,解决多人协作时 AI 编程的核心痛点,每个开发者和 AI 工具看到的项目信息不一致。它把原本塞在 CLAUDE.md、AGENTS.md 或 .cursorr…
Trellis 是一款面向团队的 AI 编程管理工具,通过结构化 spec、task、workflow 和 journal 解决多人协作时 AI 工具项目信息不一致的问题,支持同时管理多个 AI Agent 并自动同步项目记忆。
@Xudong07452910: 高质量开源项目推荐:《Agents Best Practices》—— 生产级AI Agent Harness设计指南 这是一个 provider-neutral Agent Skill,专为 Claude Code、Codex 等 AI…
A guide titled 'Agents Best Practices' providing a provider-neutral Agent Skill for building production-level AI Agent harnesses, designed for tools like Claude Code and Codex.
@Xudong07452910: 开源框架推荐:《Agency Agents》—— 232 位专业 AI 智能体,按职能分工,覆盖 16 个业务部门 如果你用过 Claude Code 或 Codex,可能遇到过这个问题:AI 在代码任务上很能干,但让它做前端设计、写营销…
Agency Agents 是一个开源框架,提供232个专业AI智能体覆盖16个业务部门,每个智能体具有独特个性、沟通风格和交付标准,支持Claude Code、GitHub Copilot等多种开发工具,并有社区翻译版本。
@Xudong07452910: 开源工具推荐:《Hivemind》—— 给所有 AI Coding Agent 共享一个大脑,从真实轨迹里自动提炼技能 用了好几个 AI Coding Agent 的人大概都遇过这件事:每个工具学到的东西都锁在自己的上下文里,换一个工具就…
Hivemind 是一个开源工具,让多个 AI Coding Agent(如 Claude Code、Codex、Cursor)共享一个记忆层,自动从使用轨迹中挖掘高质量模式并转化为可复用的技能文件,实现跨工具和跨团队的技能传播,显著降低 token 消耗和交互轮次。