Overview
LobeHub is a multi-agent collaboration platform for work and everyday life, providing a collaborative workshop where humans and agents evolve together. It transforms AI agents from one-time, isolated tools into long-term, evolving partners.
To support the long-term collaboration and evolving AI usage, LobeHub has built a next-generation Agent Harness. In this architecture, agents are fundamental units of interaction. Agents with diverse capabilities and roles can be orchestrated to collaborate on the unified platform. Therefore, the agents can share context and collaborate to complete complex tasks and systematic workflows, evolving from simple conversational tools that execute single-point commands into an agentic system.
Currently, LobeHub serves users and businesses across North America, Europe, and Asia-Pacific. Its solutions apply to business scenarios such as research automation, content production, community operations, consultation analysis, and internal corporate workflows. Compared to standalone conversational agents, LobeHub's competitive advantage lies in its native support for multi-agent collaboration, flexible integration of tools and knowledge bases, and a rich ecosystem driven by the open-source community. Furthermore, through continuous operation, agents accumulate experience and memory from their respective contexts, driving the continuous optimization of their capabilities and collaboration efficiency. AWS AI services used by LobeHub include: Amazon Bedrock (available for Global Regions only), Amazon Bedrock AgentCore, and others.
Opportunity
Breaking Through Limitations of the Original Framework to Achieve Unified Agent Management
Before applying agentic AI, LobeHub (formerly LobeChat) primarily ran AI agents on the client side. Due to constraints in user device capabilities and operating environments, it struggled with issues such as instability, high resource consumption, and unpredictable efficiency. Meanwhile, this mode constrains multi-agent collaboration, parallel task execution, state management, and unified O&M, hindering the scalable growth of complex business operations.
To break through the performance limitations imposed by the operating environment, LobeHub decided to build a server runtime for AI agents, migrating AI agents from the client side to the server side. Server-side running and scheduling of AI agents enables stable execution of complex tasks, tool usage, and coordinated multi-agent collaboration. This improves the overall performance and stability, while providing a foundation for future capability expansion and scalable operations.
To truly achieve these goals, LobeHub needs powerful cloud service resources. After evaluation, LobeHub chose to collaborate with Amazon Web Services (AWS) for the following reasons:
- Mature and stable cloud services: AWS provides mature and stable cloud services across multiple regions globally, offering robust cloud infrastructure for LobeHub's global expansion.
- Durable cost-effectiveness: From a long-term operational cost perspective, AWS offer significant advantages. Its enterprise-level market channels can further drive LobeHub's global business expansion.
Solution
Building a Server Runtime for Agents to Accelerate Business Expansion
Since partnering with AWS, LobeHub has used Amazon Bedrock AgentCore, models on Amazon Bedrock, and other services to satisfy their customers’ needs. Designed specifically for secure, large-scale agent creation and deployment, Amazon Bedrock AgentCore is compatible with various frameworks and models. It includes a series of components, providing substantial technical support for the online deployment of LobeHub's MCP tools:
- Cloud migration of local services: LobeHub uses the Amazon Bedrock AgentCore Runtime to migrate MCP tools to the cloud, resolving problems related to development environment deployment and configuration for international end users.
- Encapsulating built-in tools for AI Invocation: LobeHub uses the code sandbox feature of Amazon Bedrock AgentCore Code Interpreter to encapsulate built-in tools, enabling agents to access previously inaccessible tools by writing code.
In addition, LobeHub introduces the inference capabilities of Claude models through Amazon Bedrock and enables end users to call Claude APIs.
Using Amazon Bedrock AgentCore has not only significantly improved the development speed and reliability of our Agent Harness, but has also substantially reduced infrastructure maintenance costs. The Amazon Web Services Activate program and technical support from the AWS Startup team continuously fuel our product iteration and business expansion.
Arvin Xu
CEO of LobeHubOutcome
Enabling LobeHub's Transformation from "App" to "Platform"
Leveraging the cloud infrastructure provided by AWS, LobeHub has successfully upgraded from an "app" to a "platform", developing capabilities from infrastructure orchestration. The reliability and global reach of AWS cloud services will enable LobeHub to substantially broaden its future product growth potential and accelerate AI iteration to align with the evolving model ecosystem.
Furthermore, with the scalability of AWS elastic infrastructure clusters, LobeHub delivers secure runtime environments to customers and reduces infrastructure maintenance costs.
In the future, LobeHub will continue to explore more collaboration possibilities with AWS, such as using Amazon Bedrock AgentCore Runtime to build custom sandboxes, further expanding runtime capabilities, and batch migration of MCP tools.
About LobeHub
LobeHub is an AI platform that enables multi-agent collaboration. It is designed to enable AI agents to collaborate like human teams to help users complete complex tasks more efficiently. LobeHub provides solutions for agent creation and multi-agent collaboration. It is applicable to scenarios such as research automation, content production, community operations, and internal corporate workflows. As of now, LobeHub has surpassed 6 million installations and has 70,000 stars and 14,000 forks in the open-source community. Its primary users include individual agent users, product managers, operations specialists, researchers, consultants, content creators, independent developers, as well as open-source contributors, plugin and integration developers, AI toolchain developers, and technical leads of startup teams.