MCP Tool Invocation Issues with openai/fileParams

[kwwnjujlc]

Issue 1: Tool invocation displayed but MCP Server receives no request

Environment:

• MCP Server with openai/fileParams configured
• Tool requires user file access permission

Expected Behavior:

1. User triggers tool call
2. Permission dialog appears (for file access)
3. User approves
4. MCP Server receives the request
5. Widget loads after receiving response

Actual Behavior:

1. Frontend UI shows “tool is being called” status
2. No permission dialog appears
3. MCP Server receives no request
4. Widget loads prematurely (before any actual tool call)

Impact:
Widget displays loading state indefinitely, waiting for toolOutput that never arrives.

---

Issue 2: Duplicate tool calls with same file_id but different download_url

Environment:

• Same MCP Server setup
• Tool with openai/fileParams for file access

Expected Behavior:

1. Permission dialog appears
2. User approves
3. MCP Server receives one request
4. Server processes and returns result

Actual Behavior:

1. Permission dialog appears ✓
2. User approves ✓
3. MCP Server receives two requests within a short time window
   • Both requests have same file_id
   • But different download_url

Workaround Applied:
Implemented request deduplication based on (question, file_id) key to prevent duplicate processing.

Question:
Is this duplicate call intentional behavior? If so, what’s the expected handling pattern?

[riskbreaker]

Issue 1 is similar to one I have first sighted at December 8th. I see Tool invocation attempt in UI Frontend but MCP server never recevies the request, agent then responds that it cannot complete the task because its backend throws Resource not found error with available resources being listed as {“finite”:true}

Screen taken minutes ago. What can be noticed is that tool name is incorrect, it shows backend’s api_tool.call_tool invocation instead actual tool name - but this is randomly happens and even it shows correct tool name it still not sending requests to MCP Server.

Connector operations intermittently become "Resource not found" after several successful calls

For the last ~5 days, connector calls have been failing intermittently. A given operation works correctly several times in a row. After some number of successful invocations, the next call suddenly fails with no response, on MCP server side I see no requests happening. Once this issue occurs, even previously working calls to the same operation start failing in the same way, until end of the agent operation. Interrogation of the agent reveals that is getting error in response of tool invocati…