Multi-Agent System
A Multi-Agent System is a distributed computing architecture composed of multiple autonomous agents, each capable of perceiving its environment, executing subtasks independently, and cooperating through negotiation, message passing, or shared knowledge. This architecture is suited for complex, dynamic, or large-scale tasks beyond the capacity of a single agent, and features parallel processing, modular decoupling, and strong fault tolerance. It is widely applied in artificial intelligence, robotic control, and automated decision-making.
Core Facts
The theoretical roots of multi-agent systems can be traced back to distributed artificial intelligence research in the 1980s.
The core architecture of multi-agent systems typically involves an Orchestrator agent responsible for task decomposition and scheduling, along with multiple Worker agents responsible for specific implementation.
Multi-instance parallelism faces engineering challenges including code conflicts, context consistency, and inter-task dependency management.
Timeline (last 90 days)
The failure of multi-Agent systems can often be attributed to a single Agent's inadequate reasoning or tool use rather than the collaboration framework itself, so it is recommended to first get a single Agent working with tools such as LangGraph before introducing multi-Agent orchestration
多Agent系统的典型冲突包括：两个Agent同时修改同一文件、对同一API配额发起竞争性调用、以及在任务依赖链上产生循环等待。
Multi-agent systems introduce coordination overhead, additional inference costs, more failure modes, and potential disagreements, and should only be used when specialization brings measurable advantages.
Splitting tasks across multiple subagents introduces new problems such as merging sub-task results and resolving conflicts, for which the industry has no unified answer yet.
Core challenges faced by multi-agent systems include context window management boundaries, state consistency guarantees, and error propagation control.
Multi-agent systems represent the capability enterprises truly need for cost reduction and efficiency improvement, capable of replacing or assisting the automation of entire business workflows.
Existing agent identity verification schemes (such as cryptographically signed capability credentials) and sandbox isolation mechanisms provide partial guarantees, but are still a considerable distance from production-grade self-organizing collaboration.
Setting a communication budget in terms of dialogue turns or tokens for multi-agent systems can force the system to converge on decisions within a limited number of interactions, avoiding endless discussion
The mechanism of maintaining state through text in multi-agent systems causes communication volume to grow exponentially rather than linearly with task complexity
By pre-defining the role of each agent (e.g., planner, executor, reviewer), unnecessary mutual probing can be reduced and communication volume lowered
6 more timeline events
All Facts (20)
The theoretical roots of multi-agent systems can be traced back to distributed artificial intelligence research in the 1980s.
90%VerifiedThe core architecture of multi-agent systems typically involves an Orchestrator agent responsible for task decomposition and scheduling, along with multiple Worker agents responsible for specific implementation.
80%VerifiedMulti-instance parallelism faces engineering challenges including code conflicts, context consistency, and inter-task dependency management.
70%VerifiedIncreasing the number of parallel multi-agent instances will cause API call costs to grow linearly or even super-linearly.
65%Unverified当前多Agent系统面临的核心技术挑战是上下文窗口与长程记忆问题，模型对早期内容的关注度会随对话深度增加而下降（注意力稀释现象）
85%Unverified该多Agent系统中每个Agent本质上是一个配置了特定系统提示（System Prompt）和工具集的语言模型实例，通过结构化的消息协议相互通信
85%Unverified业界应对长程记忆问题的方案包括外部记忆存储（如向量数据库）、结构化状态机管理和周期性上下文压缩摘要，但目前尚无完美解法
80%Unverified协作式Agent在实践中遇到的问题包括信息过载、上下文污染和协作流程设计不当导致的冲突处理机制不完善
70%Unverified业界正在探索共享向量数据库、实时代码快照同步以及基于意图的冲突仲裁机制来解决多Agent协作中的上下文一致性问题
70%UnverifiedAgents communicate with each other via message passing, shared memory, or event buses.
60%UnverifiedIn LLM-based multi-agent systems, there are three main methods of inter-Agent communication: shared memory, message passing, and blackboard systems.
60%Unverified多Agent系统中Agent之间的信息传递比人类程序员之间更加完整和结构化
60%UnverifiedThe randomness of LLM outputs in multi-Agent systems causes errors to amplify layer by layer as multiple Agents execute in series; this phenomenon is known as 'Error Propagation'.
60%UnverifiedThe failure of multi-Agent systems can often be attributed to a single Agent's inadequate reasoning or tool use rather than the collaboration framework itself, so it is recommended to first get a single Agent working with tools such as LangGraph before introducing multi-Agent orchestration
50%Unverified多Agent系统的典型冲突包括：两个Agent同时修改同一文件、对同一API配额发起竞争性调用、以及在任务依赖链上产生循环等待。
50%UnverifiedMulti-agent systems introduce coordination overhead, additional inference costs, more failure modes, and potential disagreements, and should only be used when specialization brings measurable advantages.
50%UnverifiedCore challenges faced by multi-agent systems include context window management boundaries, state consistency guarantees, and error propagation control.
50%UnverifiedSplitting tasks across multiple subagents introduces new problems such as merging sub-task results and resolving conflicts, for which the industry has no unified answer yet.
50%UnverifiedMulti-agent systems represent the capability enterprises truly need for cost reduction and efficiency improvement, capable of replacing or assisting the automation of entire business workflows.
50%UnverifiedExisting agent identity verification schemes (such as cryptographically signed capability credentials) and sandbox isolation mechanisms provide partial guarantees, but are still a considerable distance from production-grade self-organizing collaboration.
50%