# Beads — Memory Upgrade Layer for Coding Agents > Beads adds persistent, structured memory for coding agents via a dependency-aware graph and a CLI, keeping long-horizon tasks on track across runs. ## Install Save the content below to `.claude/skills/` or append to your `CLAUDE.md`: ## Quick Use 1. Install / run: ```bash curl -fsSL https://raw.githubusercontent.com/gastownhall/beads/main/scripts/install.sh | bash ``` 2. Start / smoke test: ```bash cd your-project && bd init && bd prime ``` 3. Verify: - Create one issue node, claim it, update it once, then close it; confirm state persists across a new terminal session. ## Intro Beads adds persistent, structured memory for coding agents via a dependency-aware graph and a CLI, keeping long-horizon tasks on track across runs. - **Best for:** Coding agents that need durable task state (issues, dependencies, claims) instead of losing context between runs - **Works with:** CLI-first workflow; integrates with multiple agent/IDE setups (see README) - **Setup time:** 12 minutes ## How to Use It Well - Setup time ~12 minutes (install CLI + `bd init` in one repo) - Persistent graph beats markdown plans for long-horizon work (dependencies, status, claims) - GitHub stars + forks (verified): see Source & Thanks Treat memory as a data structure, not prose. If an agent can query a graph (what’s blocked, what’s claimed, what changed), it can plan more reliably and avoid re-deriving the same context every run. ### FAQ **Q: Do I need to clone the repo into my project?** A: No. The README recommends installing the CLI system-wide and using it across repos. **Q: What’s the first command to learn?** A: `bd prime` (or equivalent) to get workflow context, then `bd show/update/close` for issue lifecycle. **Q: Is it tied to one agent?** A: No. The README mentions multiple integrations (Codex, Claude, Cursor, etc.); use the setup that matches your tool. ## Source & Thanks > Source: https://github.com/gastownhall/beads > License: MIT > GitHub stars: 23,532 · forks: 1,554 --- ## 快速使用 1. 安装 / 运行: ```bash curl -fsSL https://raw.githubusercontent.com/gastownhall/beads/main/scripts/install.sh | bash ``` 2. 启动 / 冒烟测试: ```bash cd your-project && bd init && bd prime ``` 3. 验证: - Create one issue node, claim it, update it once, then close it; confirm state persists across a new terminal session. ## 简介 Beads 用依赖图+持久存储替代松散的 markdown 计划,为 coding agent 提供结构化 issue/上下文记忆;配套 CLI 与多种 agent/IDE 集成,适合长周期任务。 - **适合谁:** 需要“可持续的任务状态”(工单、依赖、认领)而不是每次 run 重新丢上下文的 coding agent - **可搭配:** CLI 优先工作流;可与多种 agent/IDE 集成(见 README) - **准备时间:** 12 分钟 ## 实战建议 - 上手约 12 分钟(安装 CLI + 在一个仓库里 `bd init`) - 依赖图+状态比 markdown 计划更适合长周期任务(依赖/状态/认领可追踪) - GitHub stars + forks(已核验):见「来源与感谢」 把记忆当数据结构而不是散文。如果 agent 能查询一张图(阻塞点、认领状态、变更历史),规划会更可靠,也能避免每次 run 都重新推导同一份上下文。 ### FAQ **需要把仓库 clone 到项目里吗?** A: 不需要。README 建议系统级安装 CLI,然后在多个仓库复用。 **最该先学哪个命令?** A: 先用 `bd prime` 获取工作流上下文,再用 `bd show/update/close` 走完整生命周期。 **只能配某一种 agent 吗?** A: 不是。README 提到多种集成(Codex/Claude/Cursor 等);按你的工具选择对应 setup。 ## 来源与感谢 > Source: https://github.com/gastownhall/beads > License: MIT > GitHub stars: 23,532 · forks: 1,554 --- Source: https://tokrepo.com/en/workflows/beads-memory-upgrade-layer-for-coding-agents Author: Script Depot