microsoft/autogen · warning · Error

Failed to get MCP capabilities

Error message

Failed to get MCP capabilities

What it means

Raised during vector search execution when vector_fields are configured and the resolved SearchQuery has empty/missing query text. Client-side embeddings need non-empty input text to vectorize; an empty string cannot produce a meaningful vector, so the tool refuses before calling the embedding provider.

Source

Thrown at python/packages/autogen-studio/frontend/src/components/views/mcp/api.ts:194

    if (!response.ok) {
      throw new Error(data.message || "Failed to get MCP prompt");
    }

    return data;
  }

  async getCapabilities(
    serverParams: McpServerParams
  ): Promise<GetCapabilitiesResponse> {
    const response = await fetch(`${this.getBaseUrl()}/mcp/capabilities/get`, {
      method: "POST",
      headers: this.getHeaders(),
      body: JSON.stringify({ server_params: serverParams }),
    });

    const data = await response.json();
    if (!response.ok) {
      throw new Error(data.message || "Failed to get MCP capabilities");
    }

    return data;
  }

  async listTools(serverParams: McpServerParams): Promise<ListToolsResponse> {
    const response = await fetch(`${this.getBaseUrl()}/mcp/tools/list`, {
      method: "POST",
      headers: this.getHeaders(),
      body: JSON.stringify({ server_params: serverParams }),
    });

    const data = await response.json();
    if (!response.ok) {
      throw new Error(data.message || "Failed to list MCP tools");
    }

    return data;

View on GitHub (pinned to 027ecf0a37)

Solutions

1. Ensure the query string is non-empty before calling the tool when vector search is enabled.
2. If empty queries are legitimate in your flow, short-circuit them before the tool call and return empty results.
3. Debug where the empty query originates — usually upstream text processing returning ''.

Example fix

# before
await tool.run(SearchQuery(query=user_text))  # user_text == "" with vector_fields set

# after
results = (
    await tool.run(SearchQuery(query=user_text))
    if user_text and user_text.strip()
    else SearchResults(results=[], metadata={})
)

def vector_query_ok(query_text: str, vector_fields) -> bool:
    return not vector_fields or bool(query_text and query_text.strip())

try:
    results = await tool.run(SearchQuery(query=q))
except ValueError as e:
    if "cannot be empty for vector search" in str(e):
        return SearchResults(results=[], metadata={"reason": "empty vector query"})
    raise

When it happens

Trigger: Invoking run() with SearchQuery(query="") or a whitespace query while vector_fields is set on the config (including the server-side path, since the check happens before choosing client vs server vectorization).

Common situations: Constructing SearchQuery with query=None defaulting to empty; building the query from an empty extraction (e.g. empty document chunk); a dict {'query': ''} forwarded from an agent.

Related errors

• Failed to list MCP prompts
• vector_fields must contain at least one field name for vecto
• vector_fields must contain at least one field name for hybri
• vector_fields must be provided for vector search
• Failed to get login URL

AI-assisted analysis of microsoft/autogen@027ecf0a37 (2026-08-15). Data as JSON: /api/errors/b521544af2f934e2. Report an issue: GitHub.