Control Your Smart TV With a Free Local AI Agent (No Cloud Needed)
Learn how to enable developer mode on LG and Android TVs and connect a free local AI agent (Hermes + Ollama) to control them over IP.
How can AI control your smart TV?
AI agents can control a smart TV once the TV exposes a way to be reached over your home network, typically through developer mode and an IP address. A coding agent (software that connects a language model to external tools) sends commands to the TV over that connection, the same way a developer would push app updates or debug settings remotely. Pair that agent with a local model running through Ollama, and you get a private, no-subscription way to operate your TV with natural-language prompts instead of a remote.
TL;DR
- Developer mode is the entry point for both LG and Android based TVs, and it’s a built-in setting, not a hack or jailbreak.
- LG TVs need a dedicated developer mode app installed from LG’s app store, followed by login and a toggle switch.
- Android TVs use the classic seven-tap trick on the build number inside Settings > About, which opens up developer options including USB debugging.
- The TV’s IP address, found under network settings, is what the AI agent connects to, so grabbing it correctly is a required step, not optional.
- Hermes agent paired with Ollama running a Qwen3 model was the setup demonstrated, giving a fully local, free alternative to cloud AI assistants.
- A single prompt defines the whole project, instructing the agent to confirm changes, avoid deleting apps outright, and keep an undo list so the TV doesn’t get bricked.
- Anyone can customize the prompt to change what the agent is allowed to do, since there’s no licensing restriction on the approach itself.
Remy is new. The platform isn't.
Remy is the latest expression of years of platform work. Not a hastily wrapped LLM.
What does developer mode actually enable on a TV?
Developer mode is a manufacturer-provided setting meant for app testing, not an exploit. On LG TVs, it comes packaged as a separate app you install from LG’s own store. Once logged in, flipping the developer mode toggle opens up the TV for external connections. On Android TV (and by extension Samsung and other Android-based sets), it lives in a more familiar place: Settings, then About, then tapping the build number seven times, a method Android phone users will already recognize. The key additional step on Android TV is switching on USB debugging inside the newly opened developer options, since that’s what allows remote commands to actually reach the device.
Neither process requires rooting the TV or installing unofficial firmware. It’s closer to turning your TV into something a developer can talk to directly, rather than only through the remote and the on-screen menu.
Why does the AI agent need the TV’s IP address?
Once developer mode is active, the TV becomes reachable on the local network like any other device, but the agent needs to know where to send commands. That means finding the IP address under the TV’s network settings. This is the same IP address you’d use for screen mirroring or casting from a phone, just repurposed here as the address the AI agent targets when it issues instructions. Without it, the coding agent has no way to locate the TV among everything else connected to the router.
How do Hermes and Ollama fit together?
Ollama is the piece that runs the language model locally on your own machine, with no cloud API calls and no per-token billing. In the setup described, a Qwen3 model was loaded into Ollama. Hermes agent is the coding agent layer that sits on top: it takes natural-language prompts, decides what actions to take, and executes them against the connected device, in this case the TV at its known IP address.
The combination works in three layers. Ollama handles the model inference. Hermes handles turning that inference into actions. The TV, now in developer mode and reachable over IP, is the thing being acted on. None of the three pieces are proprietary to a single ecosystem, which is why this pattern generalizes beyond just TVs to other IP-addressable smart home devices.
What does the control prompt actually say?
The entire project runs on one prompt fed into Hermes agent once it’s pointed at the Ollama model. That prompt does three specific things: it tells the agent to ask for confirmation before making any change, it instructs the agent to switch apps off rather than delete them outright, and it tells the agent to maintain an undo list so changes can be reversed. These three guardrails exist specifically to prevent the agent from doing something destructive to the TV’s configuration, since an autonomous agent making changes without oversight is the main risk in this kind of setup.
Remy doesn't build the plumbing. It inherits it.
Other agents wire up auth, databases, models, and integrations from scratch every time you ask them to build something.
Remy ships with all of it from MindStudio — so every cycle goes into the app you actually want.
Because it’s just a text prompt, modifying it is trivial. Anyone comfortable editing a prompt can expand or restrict what the agent is permitted to do on the TV, whether that means limiting it to volume and app switching or opening it up to deeper settings changes.
Is this safe to try on your own TV?
The practical risk is misconfiguration rather than security exposure, assuming the TV stays on a private home network. Developer mode and USB debugging are standard manufacturer features, not jailbreak patches, so enabling them doesn’t void functionality the way custom firmware might. The bigger risk sits with the agent itself: if the guardrails in the prompt are removed or weakened, a model could make unwanted changes, like disabling apps or altering settings in a way that’s confusing to undo.
The practical fallback is a factory reset, which restores the TV to its original state if something goes wrong. That safety net is part of why this kind of experiment is low-stakes compared to, say, modifying a phone’s bootloader. Still, it’s worth treating the agent’s permissions carefully, especially if you extend the prompt to allow actions beyond app switching and basic settings.
Why use a local model instead of a cloud AI service?
Running the model through Ollama instead of a cloud API means no account, no subscription, and no data about your TV usage leaving your network. It also avoids per-request costs, since local inference only costs electricity and compute, not API fees. The tradeoff is that local models, particularly smaller ones that run comfortably on consumer hardware, generally lag behind the largest cloud models on complex reasoning tasks. For a narrow task like executing TV commands based on a well-defined prompt, that gap matters less, since the job is closer to structured command execution than open-ended reasoning.
This local-first approach also means the setup keeps working even if a cloud provider changes pricing, deprecates an API, or goes offline. The entire stack, model, agent, and TV connection, lives on hardware you control.
Frequently Asked Questions
Do I need a developer account to enable developer mode on my TV?
For LG TVs, yes, you install a developer mode app from LG’s store and log in before the toggle becomes available. For Android TV, no account is required; it’s opened up locally by tapping the build number in the About settings menu.
Does this work on Samsung TVs too?
The general pattern (developer mode, USB debugging, IP address, agent connection) applies to Android TV based sets broadly, which includes many Samsung models alongside LG and others, though exact menu locations vary by manufacturer.
What AI model do I need to run this?
The setup described uses a Qwen3 model running locally through Ollama, which is free to download and run on your own hardware, avoiding the need for a paid cloud AI subscription.
What happens if the AI agent breaks something on my TV?
Because the guiding prompt instructs the agent to switch apps off instead of deleting them and to maintain an undo list, most changes are reversible. Worst case, a factory reset restores the TV to its default configuration.
Can I change what the AI is allowed to do on my TV?
Yes. The entire control logic lives in a single editable prompt, so you can restrict or expand the agent’s permissions simply by rewriting the instructions you give it.
