Connect Grok to yasync file transfers

1. Prepare the two credentials

Create a dedicated key in API keys on any registered account, including Free. A key acts as its owner; keep its value in local environment or secret settings, outside prompts and committed files.

Your xAI API key authorizes model usage; the yasync key authorizes transfer access. Set XAI_API_KEY, YASYNC_API_KEY and XAI_MODEL in the server process. Choose an xAI model that currently supports remote MCP. Model usage is billed separately by xAI.

2. Make a read-only MCP request

Save this as list-transfers.py and run it with Python 3. It uses the standard library, sends the MCP server a Bearer header and exposes only yasync_list_transfers to the model.

import json
import os
import urllib.request

payload = {
    "model": os.environ["XAI_MODEL"],
    "input": [{"role": "user", "content":
        "List my five latest yasync transfers with their status."
    }],
    "tools": [{
        "type": "mcp",
        "server_label": "yasync",
        "server_url": "https://yasync.com/api/mcp",
        "headers": {
            "Authorization": "Bearer " + os.environ["YASYNC_API_KEY"]
        },
        "allowed_tools": ["yasync_list_transfers"]
    }]
}
request = urllib.request.Request(
    "https://api.x.ai/v1/responses",
    data=json.dumps(payload).encode(),
    headers={
        "Authorization": "Bearer " + os.environ["XAI_API_KEY"],
        "Content-Type": "application/json"
    },
    method="POST"
)
with urllib.request.urlopen(request, timeout=120) as response:
    result = json.load(response)
print(json.dumps(result, indent=2))

Inspect the response for the MCP tool result as well as the model’s answer. Tool errors need handling even when the outer model request succeeds. Keep returned transfer metadata private.

3. Extend the workflow deliberately

Add yasync_get_transfer only when you need a file-level inspection. To publish a new delivery, your application must create a draft, upload and verify files, and finalize the transfer before returning a usable share link. Persist the creation request ID so a retry reuses the same draft.

The MCP tools manage transfer metadata and return signed upload plans. They do not read your disk or accept file bodies. For local files and watch folders, use the local stdio companion or the yasync CLI in an assistant session with approved terminal and file access, or implement the direct upload flow. The CLI uses its own account login rather than the MCP API key.

For a server that already has the finished files, running the CLI with machine-readable output can be simpler than exposing each upload step to the model. Remote MCP alone cannot read a workstation’s disk.

If it does not connect

Confirm that the model supports remote MCP, the xAI request uses the expected tool configuration, and the yasync key belongs to an active registered account. A direct read-only MCP request can isolate account authentication from model integration.

This is an xAI API integration. It does not imply that every Grok consumer chat screen offers custom MCP servers or file-system access.