Local models are one of the main drivers for our installer / Desktop app for OpenClaw https://holaclaw.ai (disclaimer I am one of the founders). The smaller models are really only suitable for the most basic tasks, but if you have 32gb-64gb you can get real work done (ie complex web workflows) without third party hosted models
Hola! Co-founder here. We made HolaClaw, a free desktop app for OpenClaw on Mac. It features a 1-click install (just a regular .app that you drag into the Applications folder), secure default settings (curated set of skills, agent running inside a VM) and can optionally run local models. Looking forward to your feedback!
Forgot to mention there is also a companion cli tool (also open source) that just reads directly from your agent data directories and skips the uploading part. You can also use it to anonymize a session and share it using GitHub Gists.
Hi, one of the cofounders here! We work with a ton of different agents in our other open source project, Rover (https://endor.dev/rover). Sometimes we need to debug how the agents are behaving and we built this tool to make our life easier. It is a web interface in which you can upload sessions logs that you have exported from Claude, Codex or any other coding agent. You can easily navigate them, check the reasoning, tool calls, etc. as well as explore what subagents were doing. Everything happens local to the browser, so no data is sent to any server.
It can also be used to safely share agent session logs with others, either projects you are contributing to or to troubleshoot something. It has an optional, built-in anonymizer so no private information leaks accidentally.
We applied a lot of the technical hacks described in this article and the original one to provide a full Linux environment (including networking and mounting directories) running inside the browser. https://endor.dev/s/lamp
The secondary provided that relief. Continuing to work on the company makes perfect sense if you enjoy it, which he clearly does (with the inescapable ups and downs of every start up of course).
We do Rover, which is different from the Microsoft product but the goal is similar. I was just responding to the above comment. I agree with you, it is a pretty good project and will be taking a look. We are so early there are tons of things to learn and try.
The main use case why we developed Rover internally (and still is) was the ability to run agents in parallel. It allows us to go much faster but requires tooling around it.
Secondarily it makes it easier for everyone to share those best practices and tooling among us, but is less of an issue because we are a small team
You may not be the target user for this project then and that’s fine! They are releasing this as a research project so a business model was not probably one of the key decision points.
Not sure if you are being serious or not. That was indeed the point of the very first Linux distros and why most people use them nowadays vs the alternative.
I started using Linux before there were distros (circa 1993) and it was not a pleasant experience compared to when Slackware came out
Our “product” is a tool we developed internally and found it so useful that decided to open source it.
With full potential I refer to getting the best possible results. For example, being able to work on tasks in parallel without Claude instances interfering with each other vs , well, no doing so.
I see a Linux distro as a collection of libraries that someone puts together following best practices and conventions (ie all config files go into /etc). The similarity with this project is that Microsoft has taken a collection of tools and best practices and put them together in an easy to install package
How can it be gatekeeping when they are literally making it easier to use? The analogy is probably closer to a Linux distro. You can put everything together yourself but if someone gives you a pre integrated environment with best practices it makes it easier to get started