I have just a list of chat sessions on the web app on all my projects. The webapp is modified to launch claude code daemons (borrowed from humanlayer/codelayer) and exposes the outbound STT from the WebRTC into a chat session.
- MCP Auth is via auth0
- Webapp itself is gated by a Bearer token.
This itself gets me pretty far. I am not sure what more this is offering?
My TTS/STT models are local by Kyutai and the voice agent's LLM between STT and TTS is used to determine some basic context: e.g. what project directories, mcp servers to select and what skills to use for launching the daemons.
I spend my time trying to tuning the voice+webapp experience: i.e. how it can explain things, can it surface thinking tokens from claude tools properly etc. The sweat, blood, voice go into `/create_research -> /create_plan` loop before the `/implement_plan`. Sometimes I copy the research and paste it into chatGPT for review or comments as well.
I generally use it to explain things to me to make progress, and I often pause it and ask for drawing me a mermaid a sequence diagram for events or a block diagram showing how pieces go together.
- I only had the option to create a worktree from main. I'd like to be able to pick any starting branch. Not a big deal, I just told claude to checkout the branch I wanted as my first instruction.
- For some reason in the Android app the usual automatic capitalisation of the first letter in a sentence doesn't work. Claude probably doesn't care, but I like to type in proper sentences!
- It would be nice if the worktree names got semantic names, e.g. by running my first prompt through Haiku. Maybe that's not the order things are set up in though
This is actually a PR that I have open right now, it'll be out soon!
> automatic capitalization
We'll get this fixed
> automatic worktree names
definitely on our roadmap, probably will do that this week
- Is it possible to completely disable or not use the remote sandbox features? I would never use them and would prefer my code stays on my device.
- For those of us that are using subscriptions, does it show our remaining usage? I would hate to run out of tokens in the middle of a session.
- One feature of the CC TUI I sorely missed on mobile is the ability to look up and directly reference files via “@“. Is any functionality like this planned?
- (This likely won’t affect my decision to use the service as I’ll just put it on a company card.) $20 per month for a service that runs CC on a remote machine in a convenient matter is steep but doable. Asking that same amount for a running code on my own server seems a bit unjustified, especially since this is pricier than the cost of a Claude pro subscription. Are there any plans to offer a cheaper tier for those of us that just want to run this on our own machines?
Is it possible to completely disable or not use the remote sandbox features? I would never use them and would prefer my code stays on my device.
Yes, the remote sandbox feature is disabled by default, and you have to manually enable it for the syncing to start.
For those of us that are using subscriptions, does it show our remaining usage? I would hate to run out of tokens in the middle of a session.
Currently Omnara doesn't show your usage limits, you would have to check that at claude.ai. I'll look into add that though.
One feature of the CC TUI I sorely missed on mobile is the ability to look up and directly reference files via “@“. Is any functionality like this planned?
Yes, this exists in Omnara already!
Are there any plans to offer a cheaper tier for those of us that just want to run this on our own machines?
That's a good idea, we'll think about doing this where we don't offer sandbox + voice, and just have the messaging service.
"1. For a Linux user, you can already build such a system yourself quite trivially by getting an FTP account, mounting it locally with curlftpfs, and then using SVN or CVS on the mounted filesystem. From Windows or Mac, this FTP account could be accessed through built-in software."
Obligatory https://news.ycombinator.com/item?id=9224
Welcome to the new economy.
We try to provide more features on top as well, including (but not limited to):
* improved web interface
* worktrees
* sandboxing
* richer git management (richer diffs, checkpoints, git operations)
* preview URLs
Just shows that AI has advanced so quickly that it is easy to clone YC startups to zero.
We've heard this from other users and it's on our roadmap. The challenge is we're building features like voice coding agents and hosted sandboxes that require plaintext inputs, so we'd need two execution models. Doable, but adds complexity for our team size.
That said, it's something we're prioritizing as we grow. No promises on timing, but it's coming.
If you want to discuss specific requirements or a local-only mode, happy to chat: https://discord.gg/Dc46sYk6e3
What do you mean by syncing? Happy coder syncs sessions between all my happy coder clients. I can even see in real time how happy coder in my browser's conversations progress as well as on my phone, in parallel.
Omnara also displays realtime conversations between all Omnara clients. What I mean by syncing is syncing your conversation and code changes to a cloud sandbox, which is useful if you're using Omnara on your laptop and you close your computer (as explained in the original post). If you run your agents on a persistent cloud VM, then this is less of a value add.
I can voice chat with Happy coder.
We use https://docs.livekit.io/agents/ which runs the voice agent in the cloud (to enable the above use case, and a better experience when you're using your phone when it's off), whereas I believe happy runs a client-side voice agent.
Claude Code will help you set this up using tmux. Voice input and output personally seems like a gimmick, but OS accessibilty features should work fine for this.
"Claude Code will help you set this up using tmux" - I don't think that you can connect to tmux sessions running in Claude Code/Codex remote VMs using their mobile apps. Their mobile apps don't provide a terminal emulator afaik
Right, this requires using a terminal app.
We try to provide a decent chunk of features on top as well, including (but not limited to):
* web support
* worktrees
* sandboxing
* richer git management (richer diffs, checkpoints, git operations)
* preview URLs
My current solution is to have claude (--dangerously-skip-permissions) listen for messages in my slack DMs to myself and take action in response to those messages.
I would happily switch to something better.
Why is Omnara better?
There's also Happy, Coder/Mux, and so many others that actually started out open-source and stayed that way and I can be sure my chats are not going to a 3rd party?
And Omnara chats are stored in our DB, which is how we're able to enable our voice agent, cloud syncing, and ability to see all your chats while your machine is offline. Basically anything you see in the web and mobile clients is being proxied/persisted through our server, until you delete the chat.
Our goal is to be harness agnostic as well, so eventually we will be adding support for running OpenCode sessions in Omnara.
Although I must say that Omnara's UI looks absolutely fantastic. Well done!
I think I will only explore an alternative if the UI on the mobile is dramatically better (something to explore) but im trying my best to only pay for the model these day and avoid any other tooling subscription, and doing pretty great thus far.
cheers and congrats on the launch!
If I paste in something confidential, and Omnara suffers a breach tomorrow - will my conversation data be a part of it?
I have been pretty satisfied with it, and it’s free with unlimited sessions, so I need a good reason to switch
reliability is not big selling point given the high cost on unlimited sessions which I use quite a bunch
I was curious if it had something Hapi didn't have, but sounds like it doesn't.
Thanks for your help!
Not affiliated with that project, but have been using it for a few weeks and it blows every other 'GUI for the CLI agents' I've tried out of the water in terms of both features and just working snappily/consistently.
Also totally free, and actively being improved by the solo maintainer and an active community of contributors.
Omnara providing a tunnel for you is nice, but considering Tailscale is dead simple and free, feels hard to justify $20 a month for what looks like considerably less features than openchamber
FYI openchamber/opencode support claude/codex/copilot subscriptions as well
Even though this might be "the obvious thing", I think there is a non-obvious way to build it.
Sidenote - is this novel enough to be backed by YC? Just seems like a feature that Anthropic/OpenAI could release any day.
And for your sidenote, we'll see when Anthropic/OpenAI release something similar. For now it's pretty useful, and we're making it better every day.
It's interesting to me that YC has managed to dilute this prestige to a large extent. I don't think it's an inevitable result of scale: look at Google. I think "Xoogler" prestige has diminished, but it's not nearly as bad as what has happened to the YC label.
My theory is: YC never figured out their formula. The whole formula is essentially Paul Graham, who had a knack for trusting his gut (and sometimes his wife) when everything else in his "system" was saying NO.
Once they lost that, they had to rely on what was left, and it simply wasn't competitive anymore. It's like Apple in their John Sculley era. While Sculley is credited with growing Apple's revenue from $800 million to $8 billion, his approach created a "mess of dull SKUs" that eventually confused customers and diluted the brand.
They also have a (bad) habit of removing access to bookface for all the founders who aren't "active", decimating their network and in some ways discarding valuable knowledge around what didn't work.