我们相信，记忆是智能的基石。StepMind 社区致力于构建 AI 时代的记忆基础设施—— 通过 AMCP 协议， 让 AI 的知识不再碎片化，而是可共享、可演化、跨平台、跨生命周期地持久积累。 Memory is the foundation of intelligence. StepMind builds the memory infrastructure for the AI era — through AMCP Protocol, making AI knowledge persistent, shareable, evolvable, and consistent across any platform and lifetime.
让 AI 成为 真正拥有记忆的存在——
不是每次对话归零的工具，
而是持续积累、不断进化的智慧伙伴。 Make AI a being that truly remembers —
not a tool that resets every conversation,
but a wisdom companion that accumulates and evolves.
— StepMind 社区宣言
— StepMind Community Manifesto
记忆是智能的基石。人类花了数十年才建立起今天的知识积累——而当前的 AI，每一次对话都从同一个起点出发，无法真正"成长"。
Memory is the foundation of intelligence. Humans spend decades building knowledge — yet today's AI starts from zero in every conversation, never truly growing.
StepMind 不是要构建又一个 AI 应用，而是要构建 AI 记忆层的"TCP/IP"——一个开放、统一、可进化的协议标准，让任意 AI 与任意存储之间的知识通信成为基础设施，而不是每家重复造轮子。
StepMind doesn't build another AI app. We build the "TCP/IP" of AI memory — an open, unified, evolvable protocol standard, making knowledge communication between any AI and any storage infrastructure, not reinvented wheels.
就像 HTTP 改变了信息的流动方式，AMCP 将改变 知识的积累方式。这不是一个公司的产品，而是一个时代的基础设施——由社区共同定义、共同维护、共同演化。
Just as HTTP changed how information flows, AMCP changes how knowledge accumulates. This is not a company's product — it's an era's infrastructure, defined, maintained, and evolved by the community together.
七层完整协议栈，从二进制帧到语义应用，每层有独立的刚性规范。任意 AI × 任意数据库，一次接入，永久互操作。 A complete 7-layer protocol stack, from binary frames to semantic applications, each layer with its own rigorous spec. Any AI × Any DB — connect once, interoperate forever.
一次接入，任意组合。实现一套 AMCP 客户端，即可与任意符合协议的 AI 模型和存储后端通信。不再为每个 AI 厂商写不同的适配层，不再为每个向量数据库维护不同的 SDK。
Connect once, mix freely. Implement one AMCP client and communicate with any protocol-compliant AI model and storage backend. No more vendor-specific adapters. No more managing multiple SDKs.
AMQL 查询语言类型化、可验证，比自然语言 prompt 更稳定；AMF 知识格式自带版本和因果关系，不再担心记忆污染和数据漂移。
AMQL is typed and verifiable — more stable than natural language prompts. AMF knowledge format includes versioning and causality built-in, so no more memory pollution or data drift concerns.
打破数据孤岛，建立企业知识资产。每个 AI 对话产生的知识不再随会话结束而消失，而是沉淀为企业的长期智力资本——可追溯、可审计、可跨系统共享。
Break data silos. Build enterprise knowledge assets. Knowledge from every AI conversation no longer vanishes after the session — it becomes long-term intellectual capital: traceable, auditable, shareable across systems.
分布式存储层支持端侧部署，敏感知识永远不离开企业网络；联邦同步机制允许跨部门、跨子公司的知识流动，同时保证数据主权。
Distributed storage supports edge deployment — sensitive knowledge never leaves your network. Federated sync enables cross-department knowledge flow while guaranteeing data sovereignty.
开放协议，开放研究。AMCP 的每一层都是可替换的接口，研究者可以在 L6 替换 AMF 格式探索新的知识表示，可以在 L4 替换 CSE 算法探索新的检索策略，可以在 L5 替换共识机制探索新的一致性模型。
Open protocol, open research. Every AMCP layer is a replaceable interface. Swap L6's AMF to explore new knowledge representations. Replace L4's CSE algorithm to try new retrieval strategies. Experiment with L5 consensus mechanisms for new consistency models.
所有实验结果可以回馈社区，成为下一个协议版本的技术基础。你的研究不会停留在论文里，而是真正落入生产级基础设施。
All experimental results can be contributed back to the community, becoming the technical foundation for the next protocol version. Your research doesn't stay in papers — it enters production infrastructure.
AI 终于有了真正的记忆。不再需要每次对话重新解释你的偏好、你的项目、你的决策历史。基于 AMCP 的 AI 工具能够真正记住，真正积累，真正成为你的长期伙伴。
AI finally truly remembers. No more re-explaining your preferences, projects, and decision history in every conversation. AMCP-based AI tools genuinely remember, genuinely accumulate, genuinely become your long-term partner.
而且这些记忆不被任何单一厂商垄断——它们属于你，可以在不同 AI 工具之间迁移，永远不会因为换了一个 AI 应用而失去。
And these memories aren't monopolized by any single vendor — they belong to you, portable across different AI tools. You'll never lose them just because you switched AI applications.
StepMind 采用分层治理模型，兼顾效率与开放。无论你是第一次接触 AI 协议，还是在构建分布式系统，都有你的位置。 StepMind uses a layered governance model balancing efficiency with openness. Whether you're new to AI protocols or building distributed systems, there's a place for you.
good first issue 的任务开始good first issue to startAMCP 协议的决策过程完全透明。重大变更通过 RFC 流程公开讨论，所有会议纪要和投票结果存档公开，任何人都可以查阅。 AMCP's decision-making is fully transparent. Major changes go through open RFC process, all meeting notes and voting results are archived publicly for anyone to review.
StepMind 社区正在构建 AI 时代最重要的基础设施之一。无论你是工程师、研究者、产品人还是 AI 爱好者，你的每一步贡献都会成为这个协议进化历史的一部分，永久记录在不可变的事件日志中。 StepMind community is building one of the most important infrastructure pieces of the AI era. Whether you're an engineer, researcher, product person, or AI enthusiast — every contribution you make becomes part of this protocol's evolution history, permanently recorded in the immutable event log.