Manus Launches Wide Research, Enabling Agent-To-Agent Collaboration With Hundreds Of Subagents

In Brief

Manus has launched Wide Research, a new AI-powered feature that enables flexible, large-scale parallel processing through collaborative subagents, making complex computational tasks more accessible beyond technical users.

AI company focused on agentic systems, Manus introduced a new feature called Manus Wide Research. The company states that access to advanced cloud computing resources for handling intricate workflows has historically been reserved for technically skilled users, particularly engineers. With Wide Research, Manus aims to broaden this access, offering a user interface powered by AI agents that simplifies engagement with complex computational tasks.

Wide Research is designed to support the execution of large-scale operations that involve processing and analyzing information across hundreds of items. Its intended use cases range from examining the Fortune 500 to comparing global MBA programs and evaluating generative AI platforms. The system is described as a method for conducting high-volume, in-depth research with greater ease.

While referred to as a research tool, Wide Research is positioned as a broader framework for parallelized computing and inter-agent collaboration. It departs from conventional multi-agent setups that rely on narrowly defined roles such as ‘manager’ or ‘developer.’ Instead, each sub-agent in the Wide Research system operates as a full-scale, general-purpose instance of the Manus agent. This design choice is intended to enhance task versatility and reduce the constraints associated with fixed operational domains.

The platform emphasizes flexibility, allowing tasks to be executed in open-ended formats. The feature is being released initially to Pro-tier users, with access for Plus and Basic-tier users to follow gradually. According to Manus, the launch of Wide Research represents a key point in its ongoing research into the scalability of AI agents, and the underlying infrastructure is part of a larger initiative currently under development.

Manus AI Sets New Standard In General-Purpose Agents With Autonomous Execution And Benchmark-Leading Performance

Manus AI is a general-purpose AI agent developed to convert user-defined objectives into executable actions. In standardized GAIA benchmark evaluations, the system has demonstrated leading performance across all measured difficulty levels, outperforming other AI assistants currently available.

Unlike systems limited to offering recommendations or suggestions, Manus AI is designed to deliver complete outcomes for assigned tasks. Human performance on GAIA assessments averages around 92%, while GPT-4 equipped with plugins has been reported to score approximately 15%. Manus AI significantly surpasses these benchmarks, indicating its advanced operational capabilities.

The core features that distinguish this system from comparable solutions include its structured tool interaction, autonomous operational environment, and advanced task planning capabilities. Unlike conventional chatbot systems that are generally limited to producing textual responses or offering recommendations, this AI agent is engineered to design and implement full task workflows. It does so by autonomously selecting and engaging appropriate digital tools, enabling the completion of complex, multi-step tasks from beginning to end.

Additionally, Manus AI is supported by its own independent computing environment, allowing it to function without continuous user oversight. This enables it to manage workloads autonomously, maintaining productivity even when user input is not present.

In terms of collaboration, the system facilitates a form of interaction that more closely resembles working alongside a human colleague rather than utilizing a standard software tool. It can interpret objectives, organize necessary steps, and execute processes while maintaining a communicative and adaptable interface throughout the task lifecycle.

Manus AI exists within a broader landscape of agentic AI technologies. Other developments in this space include TARS Agent, an open-source multimodal client from ByteDance focused on workflow management and real-time content generation; OpenManus, a general AI agent framework that allows for modular capability configuration and seamless tool integration; and OWL, a collaboration framework based on CAMEL-AI that also performs strongly in GAIA benchmarks.

These projects collectively point to a growing trend in the AI sector toward building more capable, self-directed agents. Within this emerging ecosystem, Manus AI occupies a prominent position for its combination of technical performance, operational autonomy, and interaction design.