The gap I'm filling: MCP handles tools, A2A handles agent discovery, but neither lets agents specify resource requirements.

With MAPLE, an agent can say: "I need 4GB GPU, 8GB RAM, complete within 30 seconds."

Key features: - Resource-aware messaging (CPU, GPU, memory, network, deadlines) - Link Identification Mechanism (LIM) for secure agent-to-agent communication - Distributed state synchronization - Type-safe error handling - Works with any LLM/framework

900+ downloads on PyPI.

pip install maple-oss

GitHub: https://github.com/maheshvaikri-code/maple-oss

Looking for feedback from anyone building multi-agent systems.