Open-Source Multi-Agent Framework Lets Teams Mix Claude, GPT, and Other Models
Open Multi-Agent, a new open-source framework, enables multi-model agent collaboration through a shared message bus and memory system.
A new open-source framework called Open Multi-Agent has been published on GitHub, offering a model-agnostic architecture for building collaborative AI agent systems. The framework allows developers to wire together agents running on different LLMs — including Claude, GPT, and any provider accessible via a base URL — and coordinate them through a shared message bus and memory layer. Each agent can be assigned distinct roles, tools, and backing models.
The project arrives as enterprise interest in multi-agent architectures continues to accelerate, with teams looking for ways to avoid vendor lock-in while still leveraging the strengths of different model families. A model-agnostic approach lets developers assign specialized tasks to whichever LLM handles them best — using one model for code generation and another for planning, for example — without rebuilding orchestration logic. The shared memory and message bus design also addresses one of the persistent challenges in multi-agent systems: maintaining coherent state as agents hand off work to each other.