Launching today
Osaurus
Open source agents that run 100% locally on your Mac
734 followers
Open source agents that run 100% locally on your Mac
734 followers
The native macOS harness for AI agents. Any model, persistent memory, autonomous execution, cryptographic identity. Built in Swift. Fully offline. Open source.
The approval gate is nice. One action, easy call. Wondering once an agent chains 15 steps to build a deck or deploy a site, does every step stop and wait for me? Fantastic launch!
Osaurus
@artstavenka1 Thank you so much! You can always "Always Allow" on some of those actions and it won't ask you again. I would leave the destructive ones to "Always Ask" just in case!
DiffSense
How good it the macuse skill? Can the mac be controlled via local LLM? or does that require better models only available via cloud? If local LLM is possible with mac-use skill, what requirment would support local mac-use skill? Btw looooooove the Laundromat Afternoon theme 😸
Osaurus
@conduit_design Great question! We run our own evaluation suite for each local model using community sourced compute. I have to say, it's decent enough to use, some models are better at it than others. Strange though, if it's good at computer use, then it's lacking in agentic coding, and vice versa. You might need to switch between two models to have best of both worlds.
https://github.com/osaurus-ai/osaurus/blob/main/reports/COMPATIBILITY.md
DiffSense
I see you also made apple script models. What was the thought process there? I know many similar apps use applescript just as an escape hatch for terminal calls.
Osaurus
@conduit_design We built Computer Use first, and it was great for what it is (clicking around the screen). We wanted to push automation a bit further, and came up with a solution to run AppleScript using local models. Since AppleScript allows chaining apps, we thought it will produce a very rich experience in automating your Mac. It's a little bit too powerful though, so be careful. It emptied my trash bin one time and I got scared.
You can read about our process here: https://osaurus.ai/blog/applescript-models
DiffSense
@tpae I guess gurdrails and escalation etc for some actions could be a powerful layer to add in the future? like different categories of actions. which wold justfiy applescript approche vs just using applescript for cli, like pplx computer does.
Osaurus
@conduit_design Yup exactly. We're also thinking about baking in the safety on the models themselves so it will short circuit it if it writes something dangerous
DiffSense
@tpae I think safety must be deterministic. The pain is already felt with the current LLMs. if your fixing an encyption issue on github you often get abrupt action not allowed in the LLMs with no explinations. Probably some security fuse went of for no reason. Other times its works fine. But its like okay, enough with the hand holding please. 😅 I think AFM also has some issue with that for some things you cant do.
Congratulations on the launch! Being offline and private is something that makes Osaurus really stand out.
I'm a bit of a dinosaur myself and always wonder - would educational content be useful for attracting users like me? A bit old-fashioned, not using agents yet, and not sure how to apply them to real use cases.
And speaking of dinos - I love the character design and animation on your website! Which dino is your favourite? Is it the same one you loved as a kid?
It could also be really fun to have a YouTube channel where dinos explain how to use agents in an easy and playful way.
Osaurus
@julia_shtogren Thank you for your kind words!
We're in the process of building out YouTube content for new users like you, and our goal is to make it simple enough for you to get started without needing complicated tutorials. Happy to answer any questions in our Discord if you need additional help.
I love our green dinosaur (his name is Dinoki), and I recently rediscovered my love for dinosaurs once my son got into it.
@tpae Great! Joined your Discord to keep an eye on updates!
Yeah, Dinoki is cool! And it's great that you have a whole dino team with different characters :) They're such fascinating creatures with such interesting shapes - lots of space for imagination. I'm a huge fan of dinosaurs and once did all 36 days of a type challenge based on them. And honestly - it was your logo that caught my eye and made me read more about the product :)
Osaurus
@julia_shtogren Welcome aboard! We have something special planned with the dinosaurs, so stay tuned. I'm sure you will love what we're about to do.
How do you plan to handle model updates and ensure the agents stay current with the latest AI advancements without requiring a full app update?
Osaurus
@aymnart We do have auto update enabled, and we do daily updates. Our builds are fully standalone, it doesn't require any servers and you can run it locally offline. You can choose not to upgrade, but your app will fetch our latest models catalog from our hugging face.
We quantize our own models and also tune our evaluations every time a new model is dropped. We've been one of the pioneers in releasing quants for smaller models. We're one of the few that is supporting range of models including the latest Hy3.
no python underneath, built your own swift mlx runtime instead — same constraint we're under for voice checkins, nothing leaves the phone. does the approval gate get old once you're chaining multi-step actions, or can you pre-approve a trusted flow?
Osaurus
@sabber_ahamed You can absolutely pre-approve a trusted flow, there's an easy "Always Allow" button per gate.
@tpae that's clean. is "always allow" scoped per gate type or per session, so a bad actor prompt-injecting mid-chain couldn't just ask nicely and get through on an already-approved flow?
Osaurus
@sabber_ahamed It's per gate type, I would suggest setting "always allow" for ones that you know is safe to do. For destructive actions, always have the gate available.
the approval-gate-before-every-action bit is what i'd actually trust daily. same constraint on our end, on-device voice checkin, nothing leaves the phone. does the approval prompt get less frequent as it learns what you always say yes to, or is every action a fresh ask forever?
Osaurus
@sabber_ahamed You have full control over it, but you can also choose "Always Allow" which will skip next time. I would suggest for destructive actions, do not set "Always Allow"