CODESOUL / AGENT-NATIVE SYSTEMS
ROOT SYSTEMS
Plasmod 负责记住，Hypha 负责运行。两套系统共同支撑智能体的长期工作。Plasmod remembers. Hypha runs. Together they support agents over long periods of work.
持久记忆与结构化证据Persistent memory & structured evidence
Plasmod 为智能体提供跨会话的持久记忆和结构化上下文，统一保存状态、事件、证据、关系与可复用知识，使长期运行的系统能够在需要时找回正确的信息。Plasmod gives agents persistent memory and structured context across sessions. It preserves state, events, evidence, relationships and reusable knowledge so long-running systems can retrieve the right information when needed.
适合BEST FOR需要持久状态、可追溯证据和可复用上下文的长期运行智能体。Long-running agents that need persistent state, traceable evidence and reusable context.
运行时与工作流编排Runtime & workflow orchestration
Hypha 是用于构建和运行智能体系统、工作流与工具链的模块化框架。它在统一运行时中连接智能体、模型、工具、记忆和执行逻辑，让系统更容易组合、观察与扩展。Hypha is a modular framework for building and running agent systems, workflows and toolchains. It connects agents, models, tools, memory and execution logic in one observable, extensible runtime.
适合BEST FOR需要整合模型、工具和记忆，并保持执行过程可追踪的开发团队。Teams that need to connect models, tools and memory while keeping execution traceable.
FULL LIFECYCLE
从记忆、执行到改进与协作，每个模块解决一个明确问题。From memory and execution to learning and collaboration, each module solves one clear problem.
跨会话保存状态、证据与可复用上下文。Preserve state, evidence and reusable context across sessions.
连接模型、工具、策略与可追踪执行流程。Connect models, tools and policies in traceable execution flows.
将评估结果转化为策略与系统迭代。Turn evaluation results into better policies and systems.
让多个智能体以更低成本交换必要信息。Help multiple agents exchange only the information they need.
PRODUCT MATRIX
AgentInfer Kit推理、评估与学习Reasoning, evaluation & learning基础设施Infrastructure↗04
BEAM多智能体通信与协作Multi-agent communication & coordination基础设施Infrastructure↗05
THETA研究分析与主题发现Research analysis & topic discovery应用产品Application↗06
宽窄·orbit情绪复盘与自我观察Reflection & self-observation应用产品Application↗07
OpenLabAI 辅助实验工作空间AI-assisted experiment workspace应用产品Application↗1FDE / ENGINEERING
除产品外，我们提供两类完整的工程交付能力。Beyond products, we provide two end-to-end engineering capabilities.
从业务问题、需求契约与系统架构出发，完成工作流、工具接入、权限边界、异常恢复、评估测试与部署交付。We move from business problem and system architecture to workflows, tools, permissions, recovery, evaluation and deployment.
围绕智能体记忆、结构化证据与业务数据，提供数据模型、存储系统、检索链路和私有化部署开发。We design data models, storage, retrieval and private deployments for agent memory, structured evidence and business data.
ABOUT CODESOUL
核心团队成员来自电子科技大学等高校，具备智能体系统、数据库与 AI 产品研发背景。Our core team comes from the University of Electronic Science and Technology of China and other universities, with experience in agent systems, databases and AI product development.