ProductsMulti-sourceSelected

Replit 推出 Missions:在预算内执行目标的智能体集群框架

Why we picked it

Replit 新出的 Missions,你给个目标加预算,一群智能体自己拆任务跑完。跟 ChatGPT 那种单模型瞎跑不一样,它有预算兜底不烧 token。做独立产品找前 10 个用户的人可以直接拿“帮我找到首批 10 个客户”这种需求试试。

Replit 发布 Missions,定位为首个 agentic swarm harness,可在设定预算内完成用户给出的目标。用户只需输入如“get me my first 10 customers”这样的目标,系统会拆解并调度多个智能体协作执行,避免模型因误解目标而盲目运行、浪费 token。预算机制是该框架的核心设计,通过限定资源消耗来约束自主智能体的行为边界。

Image · Amjad Masad
Body

技术上看,Missions 属于智能体编排层:把单个 LLM 的自由代理模式改为有目标、有预算约束的集群(swarm)执行,由 harness 负责任务拆分与协调。这与 Replit 此前的 Agent 产品线衔接,运行在 Replit 的云端开发环境中,执行结果直接落到实际项目上。

Replit 创始人 Amjad Masad 称当前 LLM 在自主执行时经常误解目标、空转消耗 token,Missions 针对的就是这个痛点。对开发者和创业者来说,这种“预算内完成目标”的框架值得与 OpenAI、Anthropic 的智能体工具对比试用,尤其在获客、增长等开放式任务上。

Original · Amjad Masad

LLMs have gotten more autonomous recently but when you give them a goal they often misunderstand you, work aimlessly, and waste your tokens. Today, we’re introducing Missions—the world’s first agentic swarm harness that can accomplish any goal you give them within set budget. Try it today with something like “get me my first 10 customers:” replit.com/blog/introduci…

Industry Coordinates & Source Verification

This report is categorized under AITOP's "Products" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include Replit, Missions, agents, coding assistants.

Source attribution: curated from Amjad Masad (@amasad). Cross-source corroboration: 11 independent sources have covered this development (including 阿里云 Alibaba Cloud, lmarena.ai, marktechpost), confirming industry-wide signal strength.