I wanted to make a quick demo to show what MCP does, and how it's different from regular tool calls. I've seen wayyy too much talk about MCP on my timeline and felt there were a lot of people scratching their heads, so I wanted to demonstrate how we did this with our framework emp-agents.empyrealsdk.com/. The difficulty of integrating MCP is trivial. It's the quality of the MCP server that adds the value. Likewise, the ability to dynamically add and remove relevant MCP server prompts, tools and resources from your agents context window can have a big impact on the quality of its responses. I'd also like to emphasize, the reason I think open source MCP servers are so important is because you're either sending raw information from your conversation history to an external server, that is in turn injected back into your conversation, or you are executing arbitrary code locally. Blindly trusting an MCP integration without any kind of peer review or open sourcing and frozen versioning is pretty risky. There's a whole new type of attack vector that this could open up if people aren't cautious. So.... if you're interested in working on open source implementations of MCP servers for popular protocols, where we can actually bring simpler integrations for users looking to utilize LLMs to interact with protocols like... - AAVE - Uniswap - Pendle - GMX - LZ/wormhole/debridge/orbiter - Non-EVM blockhains like solana and sui - anything that has utility but needs a simpler, LLM oriented interface Please reach out. We started working on protocol abstractions in python for simpler machine learning integrations over 2 years ago when we broke ground on eth-packages.empyrealsdk.com…, and I think it would be for the general good of the ecosystem to have more people contributing to open source MCP servers. I believe every team should maintain an open source MCP for the happy path using their protocol. If any teams interested in working together to create one... again, please reach out.