Realtime AI News
NVIDIA App 11.0.9.565 turns on a local MCP server by default
Version 11.0.9.565 of the NVIDIA App adds a Model Context Protocol server for AI agents that is enabled by default and runs locally. It lets an agent read game and performance data, control capture, and change optimization and laptop power settings, prompting some users to switch it off.

The latest NVIDIA App update, version 11.0.9.565, adds an MCP server for AI agents that is enabled by default and runs locally, according to a HappyGamer report. MCP stands for Model Context Protocol, a standard that lets AI tools talk to other applications; here it means an AI assistant on a user's PC can interact with the NVIDIA App directly.
The report traces the discovery to a social post from user Pirat_Nation on Friday, October 16. According to that post, the server gives AI tools access to ten controls inside the NVIDIA App. Some are read-only; others actually change the user's setup.
On the read side, an AI tool can check the current game, show performance statistics and inspect the state of NVIDIA drivers. On the capture side, it can start and stop recordings, take screenshots, and control Instant Replay and Highlights. Those are actions users previously performed through the overlay or keyboard shortcuts.
The controls that change settings matter more. An agent can toggle RTX Dynamic Vibrance, find and launch games, change game optimization settings, and adjust Battery Boost and Whisper Mode on laptops. In other words, an MCP-capable agent is not limited to reading state; it can alter visual tuning and power policy on its own judgment.
The dispute is about the default. HappyGamer notes that the server runs locally rather than as a cloud service reaching into the PC, but the objection is about who made the choice. Pirat_Nation's own take in the post was blunt: turn it off.
The replies cited in the report follow a similar line. One user wrote that "enabled by default" was doing a lot of heavy lifting, adding that nobody asked for an AI agent poking around their games and recordings. Others said they had not known the setting was on, switched it off after seeing the post, and in one case found a second default they also had to disable.
That is a small sample, but it shows a pattern: people learned about the feature from a social post rather than from the app itself. For software bundled with a very large installed base of graphics cards, how defaults are set shapes how users perceive the boundary of access.
The practical advice is specific. If you run the NVIDIA App, confirm you are on version 11.0.9.565 or later, then look through the settings for the MCP server option and decide whether AI tools should touch your recordings and game settings. Users who already run an MCP-capable assistant can get value from launching a game or pulling performance stats on request; users who do not run agents gain nothing, and leaving it on only adds access they did not ask for.
The significance goes beyond one version bump. Desktop software is starting to ship agent interfaces like MCP as ordinary features, and the choice of default will determine how quickly that access spreads. What to watch next is whether NVIDIA adjusts the default or adds clearer disclosure in a later release, and whether other PC software vendors follow the same pattern.
Why it matters
A default-on agent interface exposes a local control surface to any MCP-capable assistant on the machine, even without cloud involvement. Expect pressure on desktop vendors to justify defaults and disclose agent-facing permissions more explicitly.
Nearby Updates
All10/18, 11:48
OpenAI quietly lists three Aurora Alpha models in its API without documentation
OpenAI has quietly added three Aurora Alpha models to its API without an accompanying announcement, according to Crypto Briefing. The report says the undocumented listings could make developers cautious about adopting models whose behavior and availability remain unclear.
10/18, 11:07
Chinese models draw 4.5x US token usage on OpenRouter as all top five slots go to China
Chinese AI models accounted for 86.74 trillion of the 172 trillion tokens processed on OpenRouter in the week of October 5-11, 4.5 times the US total, with all five most-used models coming from China. The report credits prices 60% to 90% below leading US models, open weights and local-language tuning for driving adoption across Southeast Asia, Africa and Western enterprises.
10/18, 10:07
Guangdong tops 195 filed large models, ranking among China's leaders
Guangdong has filed 195 large models, ranking among China's leading provinces, Sina Finance reported. The count measures models cleared for public service rather than lab projects, though the report gives no breakdown by sector, size, or usage.
10/18, 09:53
17-Year-Old Kimi Paper Contributor Reportedly Leaves Moonshot AI for the US
A 17-year-old researcher who contributed to Moonshot AI's core Kimi paper has reportedly left the company for the United States, according to HK01, which noted his work drew public praise from Elon Musk. The departure highlights both the thin pipeline of young talent in China's model labs and the pull of overseas opportunities for its most visible researchers.