Entoli, our private client-side assistant, is entering closed testing. From today, early builds land in testers’ hands on Android, and the feedback from this phase shapes everything that follows. Here is what Entoli is, what it stands for, and how to get in.
What is Entoli?
Entoli is a chat assistant with a coding agent behind it. It looks and feels like the chat apps you already know, but it can read and edit files, run commands, and carry work across conversations, on Linux desktop, Android, and the web. Entoli (εντολή) is the Greek word for a command: both the one you give and the one the agent runs.
Where other assistants process your requests on someone else’s machine, Entoli runs on yours. It is a client: it calls the model endpoints you configure, and it runs the agent’s commands on your own hardware. There is no server in the path, nothing to host, and no copy of your work anywhere you did not put it.
What Makes Entoli Different
We built Entoli around a set of principles, not a feature list:
Your keys, your endpoints, your hardware. Entoli calls the LLM provider you configure, OpenAI-style or Anthropic-compatible, directly from your device. Self-host your model, or use a provider you trust; the app takes no cut and inserts no middleman.
Sovereignty over your conversations. No analytics, no telemetry, no crash reporting. Your conversations are not training data for anyone. The only traffic the app generates is what you send yourself.
No watermarking. What your assistant produces is yours alone: the same words, with nobody else’s fingerprint stamped into them.
Everything on the device, if you want it. Entoli can run a language model entirely on your device with no network at all, process voice offline, and (on Android) run a whole Linux distribution in a sandbox the agent can work in.
Sync that belongs to you. Pairing two of your devices connects them end-to-end encrypted, peer-to-peer: the relay, if you use one, sees only ciphertext.
Open source. Entoli is developed under the AGPL-3.0, so what runs on your machine is code you can audit, fork, and run yourself.
Byte Repair Inside
Entoli also ships the Byte Repair workspace for verification-backed coding. Code the agent writes is checked with formal verification, the same mathematics that powers the rest of Byte Repair, so vulnerabilities like buffer overflows and memory leaks are found and repaired before they reach you. The loop only closes when the math agrees.
How to Join
Closed testing runs through Google Play on Android, and joining takes a few minutes:
- Join the Entoli group: groups.google.com/g/entoli
- Opt in to testing: play.google.com/apps/testing/io.byterepair.entoli
- Install from Play: play.google.com/store/apps/details?id=io.byterepair.entoli
- Give feedback in the group: groups.google.com/g/entoli/c/ma2d58WHvFg
We are ironing out the last quirks before release. Your feedback tells us what breaks and what confuses before it reaches everyone.
Join us, and help build an assistant that answers to nobody but its owner.