Having an AI assistant is all the rage these days, but AI assistants usually don’t know about your automation setups and may have difficulty dealing with tasks asynchronously. Enter zclaw. It gives you the option to have a personal assistant on an ESP32 backed by Anthropic, OpenAI, or OpenRouter. The whole thing fits in 888KB, and while it doesn’t host the LLM, it does add key capabilities to monitor and control devices connected to the ESP32.
You communicate with the assistant via telegram. You can say things like “Remember the garage sensor is on GPIO 4.” Then later you might say: “In 20 minutes, check the garage sensor and if it is high, set GPIO 5 low.” It has an RTOS for scheduling tasks and is aware of the timezone and common periods. Memory persists across reboots, and you can pick different personas.
Some of the use cases mentioned in the manual show how having something that can precisely schedule, control, or monitor devices might pay off. Ideas like bringing up a lab setup, scheduling plant watering, and more would be difficult to do with just a stock chatbot.
The AI can also introspect. For example, you could create a few tasks on a schedule and then ask the device to “show me my schedules.” You can also create up to 8 tools with a name, description, and action. This lets you describe something like “power_down_bench” and then tell zclaw to execute it on demand or even on a schedule. Overall, an interesting and well-documented setup.
We’ve seen many projects like this, and each has its own charm. And its own personality.
I swear to the spaghetti monster someone’s gotta be astroturfing these dang claws. The hype feels so fake.
You pay the cloud so many tokens to do things this way.
Cool to demonstrate how tiny the client side is though. I’ll grant that a tiny RTOS with configuration and scripting controlled by an LLM is pretty neat.
You can just host your own AI.
Most of this just seems like something you’d implement on a PLC with ladder-logic and remote management.
So where and why is there an angle to cram in the most inefficient way of processing commands into the mix?
No, seriously, this article makes zero sense.
Its a hack!
I mean yeah if you knew how to program a PLC then you’d do that, but this approach uses natural human language. Obviously not for serious stuff (yet) but it’s interesting.
Struggling to understand what part of the Ven diagram intersects “can’t code and needs human language” (especially given stuff like Scratch exists) and “can remember what’s connected to GPIO5”.
People like me who are part of the former and trying to learn how to do the latter
“on an esp32”
… no? it’s connecting to something remotely. this is no more impressive — actually, arguably less impressive — than running meshtastic on an esp32 or something.
Ha ha so ma,ny projects,
I exposed Camera and Speech throught ESP32
https://www.youtube.com/watch?v=H84T8Khg7vw
or a Minitel for instance
https://www.youtube.com/watch?v=NWMwm7-6qmw
That was fun, as soon as you have a Websocket and declare tools on LLM side
With AI we can finally create a device to open and close garage doors
Sorry but what we NEED isn’t useless ai chatbot on an esp32. We need ai to VIBE CODE WHOLE OPERATING SYSTEMS AND APP STORES for an esp32. Something to flash it. And maybe flash over the air everyday with a new os the ai creates for us. But I believe the ai can create the menus and os for people and we will all have a lot of cool software for esp32s.
There’s no other way to do it that fast. As a normal person I can’t pay to have esp32 software made but ai will help me make an esp32 tamagochi or an esp32 meshtastic game.
The AI assistant is great. The examples you are seeing are simple and basic but they will evolve into very ambitious projects. I have been developing IoT servers and esp32 software some involving multiple other MCUs for over 3 years. Category I wish I could afford to use some of this AI platforms because that is the way of the future.
What are you on about? Esp32 has FreeRTOS capability already. Microcontrollers are not for general purpose computing, they’re for specific tasks.
Micropython is the closest thing you’ll get to an “OS” that you can upload “apps” to run on the ESP and they barely have resources even for that – try enabling webREPL and see how much ESP32 is spare afterwards :)
My post was deleted by AI apparently.