maybe bug: text vs document in mcp response

[f1-outsourcing]

I am seeing mcp research response with text but I think continuedev is expecting document.

[ibondarenko1]

which MCP server are you using and what does the raw response look like? continue parses tool results per the MCP spec where each content block has a type field, and i think the two valid types are "text" and "image" (resource was added later). "document" isn't in the spec, so if your MCP server is returning {type: "document", ...}, continue is dropping it.

if your server is wrapping a search result, the canonical shape is:

{
  "content": [
    {"type": "text", "text": "result snippet here"}
  ]
}

not:

{
  "content": [
    {"type": "document", "document": "..."}
  ]
}

newer MCP versions added "resource" with uri/mimeType/text but you'd serialize that with type: "resource", still not "document".

if you're on someone else's MCP server, file with them. if it's yours, change the shape to {type: "text", text: "..."}. you can paste the raw response (run the server with --verbose or wrap in a logger) and it's quick to confirm.