✅ Deploy Preview for continuedev canceled.

Thanks @04cfb1ed I closed my PR to avoid confusion as it was just a quick draft I didn't spend much time on, and yours is more complete.

Hi @04cfb1ed,

Following @nejch's suggestion on the now-closed PR #5511, I wanted to reach out here. Thank you for picking this up and creating a more complete implementation for SSE/WebSocket MCP server support! This is a feature I was really looking forward to for my own setup.

Since the original PR was a draft and I was eager to test this out, I quickly put together a working version (for my specific SSE use case) in my fork:
https://github.com/enzoescipy/continue/tree/sse-mcp-support

It mainly involved completing the URL handling in loadYaml.ts and ensuring the schema definition was correct, building upon the existing logic in MCPConnection.ts.

Please feel free to reference it if any part might be helpful for finalizing this PR (#5517). No pressure at all, just wanted to share my work in case it's useful.

Thanks again for implementing this much-needed feature, and looking forward to seeing this merged!

@04cfb1ed waiting for your powers for this megical feature :)

All contributors have signed the CLA ✍️ ✅
_{Posted by the CLA Assistant Lite bot.}

I have read the CLA Document and I hereby sign the CLA

The minimal python SSE server to test is this:

# requirements.txt
# Server
fastapi==0.115.4
uvicorn~=0.34
# MCP
mcp[cli]~=1.8.0

from mcp.server.fastmcp import FastMCP
from mcp.server.fastmcp.prompts import base
from starlette.applications import Starlette
from starlette.routing import Host, Mount

mcp = FastMCP("My App")


# Add an addition tool
@mcp.tool()
def add(a: int, b: int) -> int:
    """Add two numbers"""
    return a + b


@mcp.prompt()
def review_code(code: str) -> str:
    """Review code prompt"""
    return f"Please review this code:\n\n{code}"


@mcp.prompt()
def debug_error(error: str) -> list[base.Message]:
    """Debug error prompt"""
    return [
        base.UserMessage("I'm seeing this error:"),
        base.UserMessage(error),
        base.AssistantMessage("I'll help debug that. What have you tried so far?"),
    ]


@mcp.resource("greeting://{name}")
def get_greeting(name: str) -> str:
    """Get a personalized greeting"""
    return f"Hello, {name}!"


# Mount the SSE server to the existing ASGI server
app = Starlette(
    routes=[
        Mount("/", app=mcp.sse_app()),
    ]
)

if __name__ == "__main__":
    import uvicorn

    uvicorn.run(
        "minimal:app", host="127.0.0.1", port=8000, log_level="debug", reload=True
    )

mcpServers:
  - name: Debug Local Minimal MCP Python
    type: sse
    url: http://127.0.0.1:8000/sse

Result

You folks saved my day!!

Looking forward to testing this together with https://github.com/automation-ai-labs/mcp-link 🤩

Great, I can't wait for it to be merged and available to use!

@Patrick-Erichsen
I would like to ask about this SSE PR. Thank you

After this PR is completed, is it supported to update the information returned by the tool in real-time in the UI?

The scenario I'm currently facing is: I have an MCP tool that always running for a long time, so I need to let the user see the running status in the UI.

If it is SSE protocol communication, I believe it should be able to stream back the current running progress information of the tool, such as running step 1... running step 2... status, displayed in real-time in the UI, so that users can know the execution status of the tool

If my SSE MCP tools use a streaming API to continuously return chuck data, I hope it can be updated in the UI

Thanks for the great PR @04cfb1ed !

@AfterStories - that definitely sounds possible. Is it supported by MCP protocol currently? If you don't mind, please open a new issue and let's move the conversation there to keep this PR discussion focused.

Hey @04cfb1ed , apologies for missing this in my original review but there appears to be an issue with configs that don't specify the transport type, so I had to rollback this PR.

The following config.yaml is failing:

mcpServers:
  - name: Memory
    command: npx
    args:
      - "-y"
      - "@modelcontextprotocol/server-memory"

with this error: Unsupported transport type: undefined

There also appears to be a typecheck error here: https://github.com/continuedev/continue/blob/main/core/config/yaml/loadYaml.ts#L454-L466

I don't think I/the team will have the bandwidth this week to look into it, but @04cfb1ed if you or anyone else can pick it up it would still be awesome to get SSE support merged 👍