You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
An MCP server with tool-calling support and context filling will be cool. It will complement the vdb MCP, which was released recently.
However, from my testing, tokens are consumed very quickly, as Claude appears to count tokens for the entire message (including the BOM), even if generated by the tool. The absence of a powerful local model (where is cdx1?) means that things are not entirely private.
Perhaps organizations and teams that have already adopted an MCP-compatible service would find this feature interesting?
The text was updated successfully, but these errors were encountered:
An MCP server with tool-calling support and context filling will be cool. It will complement the vdb MCP, which was released recently.
However, from my testing, tokens are consumed very quickly, as Claude appears to count tokens for the entire message (including the BOM), even if generated by the tool. The absence of a powerful local model (where is cdx1?) means that things are not entirely private.
Perhaps organizations and teams that have already adopted an MCP-compatible service would find this feature interesting?
The text was updated successfully, but these errors were encountered: