On 26th Nov 2024, Anthropic introduced the world to MCP - Model Context Protocol. Four months later, OpenAI announced the adoption of MCP across its products, making the protocol the de facto standard of the industry. Put it another way, a few months ago, we were figuring out when something is a workflow and when it is an agent (and we still are). Today, the question is how much MCP Kool-Aid we should drink.
I think MCP is really great and can be seen as "USB-C port for AI applications". Personally I can now build a quick MVP from making PRD to design to actual implementation. The only concern for mass usage in a company is mostly security I believe. However, it's actually pretty straight-forward nowadays to implement own MCP servers so companies can just create and use their inhouse MCP servers to curb the security issues.
The last time IBM was a major force in the tech scene was the release of Watson, it was a while ago. I don't expect them to make a comeback unless their research on quantum computing yields results. I would bet on MCP extending its scope to cover agent collab; or Google's recent A2A protocol. And because the distinction of tools (MCP) and agents (A2A) is increasingly blurry, there will be a tug of war between the giants. They will co-exist for now however.
I think MCP itself is the next best thing that came in AI space. I couldn't help myself playing around with my Claude Desktop and Cursor and amazed how much it can do. Once MCP is applied massively, AI will get more interesting with more use cases!
Yes. Especially now that agent-to-agent collaboration is not yielding the expected results (yet), a smart model wilth numerous tools would be the first to transform the workspace.
I think MCP is really great and can be seen as "USB-C port for AI applications". Personally I can now build a quick MVP from making PRD to design to actual implementation. The only concern for mass usage in a company is mostly security I believe. However, it's actually pretty straight-forward nowadays to implement own MCP servers so companies can just create and use their inhouse MCP servers to curb the security issues.
Regarding the hybrid system mentioning in the article, I think IBM is onto something similar so most likely we'll see it soon: https://research.ibm.com/blog/multiagent-bee-ai
The last time IBM was a major force in the tech scene was the release of Watson, it was a while ago. I don't expect them to make a comeback unless their research on quantum computing yields results. I would bet on MCP extending its scope to cover agent collab; or Google's recent A2A protocol. And because the distinction of tools (MCP) and agents (A2A) is increasingly blurry, there will be a tug of war between the giants. They will co-exist for now however.
I think MCP itself is the next best thing that came in AI space. I couldn't help myself playing around with my Claude Desktop and Cursor and amazed how much it can do. Once MCP is applied massively, AI will get more interesting with more use cases!
Yes. Especially now that agent-to-agent collaboration is not yielding the expected results (yet), a smart model wilth numerous tools would be the first to transform the workspace.