Can DeepSeek V4.1 Flash call Sume's hosted MCP tools?

Yes, through your own client code. The vendor's docs list tool calls as supported for DeepSeek V4.1 Flash, but they do not speak to Sume. You write the loop: read tools_list from Sume's hosted MCP, give the model the tool names and schemas as its function list, execute the call it picks against Sume with your credentials, and return the result. The key stays in your process and never goes into a prompt.

Who fills which field

Paid tools need the same care here as anywhere. Add dry_run, idempotency_key and max_spend_usd in your code, not in the model's arguments.

def finalize_args(model_args: dict, job_id: str, cap_usd: float) -> dict:
    args = {k: v for k, v in model_args.items()
            if k not in ("idempotency_key", "max_spend_usd")}
    args["idempotency_key"] = f"{job_id}-v1"
    args["max_spend_usd"] = cap_usd
    return args


print(finalize_args({"text": "hello", "max_spend_usd": 999}, "job-7", 0.5))

Start read-only

Under a read-only OAuth scope the paid tools are not in the list at all, which is a cheap way to run a first trial of a new model with no spend path.

Check the docs before you ship

Sume's limits and field names change faster than blog posts do. Read the linked docs pages for the current request fields before you ship, and send a dry_run or a low spend cap on your first real call.