|
|
|
|
|
by luigipederzani
203 days ago
|
|
One thing I’ve been thinking about:
Should Code Mode live purely in the MCP client, or could there be real value in having MCP servers also expose a “code mode” execution surface themselves? For some mcp servers, having domain-specific batch operations or embedded execution could unlock more efficient workflows. |
|
We have a platform (Gram) which is serverless MCP. Now users can launch MCP servers that only have 3 tools, search, describe, execute. Gram handles the tool routing, and execution (acting sort of like the sandbox). So the client only ever needs to know about those three tools. Everything else is abstracted.