Add simple streaming tool in FastMCP during tools usage #114
Description
I’ve been experimenting with a simple streaming tool in FastMCP:
@mcp.tool()
async def process_data_with_updates(data_uri: str, ctx: Context) -> dict:
await ctx.info(f"Starting to process data from {data_uri}...")
await ctx.report_progress(progress=0, total=100)
for i in range(5):
await asyncio.sleep(2)
await ctx.info(f"Step {i+1}/5 complete")
await ctx.report_progress(progress=(i+1)*20, total=100)
await ctx.info("Done!")
return {"status": "success"}Expected behavior
While the tool runs, I should see intermediate ctx.info(...) log lines and/or ctx.report_progress(...) updates appear in the LLM chat interface.
After completion, the final return value should be displayed.
Actual behavior
The tool hangs until the end.
Only the final return value is shown.
No progress updates or info logs appear inline.
Not sure if the LLM notes down the intermediate updates internally.
Notes
The chat interface is not displaying these real-time updates. This could be due to:
- The client not supporting progress tokens
- The streaming updates being sent but not rendered in the UI
- The updates being processed but not shown to the user
Would be great if LLM clients supported showing live updates in the conversation window.
Environment
- fastmcp version: 2.12.3
- Python version: 3.12
- Transport: streamable-http