Is it free?
Is it free?
Yes. The extension itself is free and open source (Apache-2.0), now and always. On top of that, the free tier lets you start with no key, no account, and no card: it runs on a managed model out of the box. When you want a specific model or heavier usage, bring your own OpenAI-compatible key.
What data leaves my machine?
What data leaves my machine?
Your prompts and the context the agent needs go to whichever model provider you use. On the free tier that is our managed model; with your own key it is your provider. See privacy and security for the full picture.
Which models can I use?
Which models can I use?
Any model on an OpenAI-compatible endpoint with strong native tool-calling. Start with Claude Haiku 4.5 (the benchmark model). See Models for recommendations and exact settings.
Can I run it fully local or offline?
Can I run it fully local or offline?
Yes, point it at a local OpenAI-compatible server such as Ollama or LM Studio. The model still needs solid tool-calling to drive hardware workflows.
Which boards and SoCs are supported?
Which boards and SoCs are supported?
Nordic nRF52, nRF53, and nRF54 on nRF Connect SDK (Zephyr), and Espressif ESP32, ESP32-S3, and ESP32-C6 on ESP-IDF. BLE on both, Wi-Fi on ESP. More is on the roadmap.
Do I need the vendor tools installed?
Do I need the vendor tools installed?
Yes. Adsum drives your existing toolchain: the nRF Connect Extension Pack for nRF, and an ESP-IDF install for ESP. See getting started for setup.
Was my code used to train it?
Was my code used to train it?
No. The expertise behind the agent is firmware knowledge authored and curated by engineers, not scraped from user projects. See how it works.
Is it affiliated with Nordic or Espressif?
Is it affiliated with Nordic or Espressif?
No. Adsum IoT Coder is an independent open-source project. It works with their SDKs but is not endorsed by or affiliated with either company.
Where are the release notes?
Where are the release notes?
In the changelog on GitHub.

