Firstly, thanks for creating this! It looks very featureful, fast, and promising.
ACP
I, and others, might like to use this with editors like Zed.
I'm not 100% postive ACP makes sense for your scope/use, etc.
I'm also not 100% postive ACP makes sense as a separat(ed) thing vs. MCP.
I am well aware of how they are (purported to be) different in terms of design and scope. But I also feel that MCP already fundamentally blurs the lines of scope:
E.g.
A "resource" really is a very simple version of a tool: "Get me this data" still needs to do some processing to make sure useful formats and forms are loaded.
A "prompt" really is just a form of algorithmic information, just like a tool is, in that it encodes how to do something.
MCP maybe could just be extended to consider an editor like Zed as a tool for getting information from loaded files, maybe via it's tree-sitter graph, and asking about current user intent.
Anyway. Just a thought. I'm not sure if ACP will take off in terms of adoption.
Also not sure how easy it would be to implement here, even if it becomes a loser in the "competing standards competition".
Thanks for your time and effort.
Firstly, thanks for creating this! It looks very featureful, fast, and promising.
ACP
I, and others, might like to use this with editors like Zed.
I'm not 100% postive ACP makes sense for your scope/use, etc.
I'm also not 100% postive ACP makes sense as a separat(ed) thing vs. MCP.
I am well aware of how they are (purported to be) different in terms of design and scope. But I also feel that MCP already fundamentally blurs the lines of scope:
E.g.
A "resource" really is a very simple version of a tool: "Get me this data" still needs to do some processing to make sure useful formats and forms are loaded.
A "prompt" really is just a form of algorithmic information, just like a tool is, in that it encodes how to do something.
MCP maybe could just be extended to consider an editor like Zed as a tool for getting information from loaded files, maybe via it's tree-sitter graph, and asking about current user intent.
Anyway. Just a thought. I'm not sure if ACP will take off in terms of adoption.
Also not sure how easy it would be to implement here, even if it becomes a loser in the "competing standards competition".
Thanks for your time and effort.