mo

Paseo is a cross-device app to manage your agents. Self-hosted. Open-source. Works with your existing Claude Code and Codex setup.

right now it's manual through github, i just make sure everything i want to work on is pushed and pull on the other machine to make it easier i have a `devbox` skill where i show agents how to ssh and sync everything up with set commands I've got ideas on building cross daemon sync through Paseo but haven't got there yet

@moboudra @paseo_sh So how do you maintain parity or sync between your laptop and the Hetzner box?

cloud agent sandboxes are cool, and i’ll probably add them to @paseo_sh at some point. but i still prefer handing work off to a real VM with my dotfiles, custom skills, and full dev setup already there. i travel a lot, so when my laptop is shut i just run agents on my hetzner box from my phone. same basic benefit as cloud handoff, but in your actual environment.

@badlogicgames i got some arabic earlier today

gpt 5.4 started spitting out chinese and invalid json. guess it's time to go to bed. (yes, i'm far past the dumb zone. it's a vibeslop project, don't care)

@paseo_sh is without a doubt the most slept on orchestrator right now. I've tried everything and there's always something they're missing. @moboudra is the first person to just nail it. Open source, every OS, and a mobile experience that truly blew me away. There is seriously nothing else like it right now. I give it a month or two before it blows up.

@ceeebeeebeee @paseo_sh cheers, appreciate the shoutout! slowly getting there 🫡

this always reminds me of what happened with stable diffusion models people are still squeezing value out of them, creating images that rival frontier image gen models, the models are the same, but the tooling around keeps improving i do feel the same will happen with LLMs, even if model development stopped right now, we'd keep building tooling to squeeze as much capability as possible

feels like model providers have shifted from chasing 100% intelligence to some percentage split between raw intelligence and more-vertical products i think its a signal that making models generally smarter is topping out and its more cost-effective to fill in the gaps with product (ie context engineering and task-specific RL - claude code, cowork, etc) to me the split feels like 80/20 product/intelligence right now. Wdyt? Is this because we have models capable of AGI and now we just need to wire them up? Or is this just a symptom of hitting a wall and venture-backed businesses need to keep pumping value through other parts of the stack? How long until the bitter lesson kills this? Year? Years? a decade?

@danrobinson i mix `paseo schedule` and `paseo loop` with @paseo_sh to keep agents working until task is done

Is there a better solution to Codex laziness than this?

crossed 600 stars recently still a vanity metric, still feels good

@hiarun02 why not build the 13th competitor

> Every idea feels taken. > Every API already exists. > Every SaaS has 12 competitors. So what do we build now?

@gaanasrini pro tip just ask it to push, commit is implied 99% of the time

am i nuts for doing it this way

@jlongster i disable memory features in everything i use for this exact reason. i find llms are not great at picking what constitutes a "memory" worth remembering

regarding agent memory, I'm realizing: I never want anything loaded automatically. no loading yesterday's memories etc (AGENTS.md is different, not memory) I don't want chronological memories. I want topical memories grouped based on what I'm doing I want to be explicit about it generally, especially saving but also reading. some automatic reading based on me switching "into" a topic might make sense most thing I see don't really fit what I want

woah this is really cool to see would love to have a good plugin system to make this stuff possible

@jasonkneen wow, really cool. i'd definitely be open to merging the browser pane upstream since that's on the roadmap, just not the canvas stuff yet, it's a much bigger shift not sure how coupled they are though

@moboudra I can do a PR today for the browser and canvas feature if you don't mind them together

paseo🤝pi @badlogicgames

@jasonkneen what's the canvas?

@moboudra how do we feel about browsers :)

@jasonkneen hundred times yes! you beat me to it :)

@odd_joel great work with Moshi by the way, looks great.

@moboudra really cool to see voice-first coding taking off. we went a similar route with Moshi - on-device Whisper for voice-to-terminal on iOS. no cloud round-trip so it stays snappy even on spotty connections. curious how Kokoro handles the latency for real-time dictation?

yes! paseo allows you to have a voice chat with Claude Code & Codex, in mobile and desktop it's all done locally via Parakeet and Kokoro models it needs some polishing, but it works code hands free :)

agreed, would really love for fully hand off voice coding to be usable, Paseo actually started voice only but scope creep lead me here :) I benchmarked most models supported by sherpa-onnx and ended up with Parakeet for STT and Kokoro for TTS Parakeet is really fast for dictation and I found it comparable to Whisper, smaller too. Kokoro synth is not super fast but it's hands down the best quality voice right now.

@0xbhargav hey Bhargav, Pi supports logging into your max sub

@moboudra Can I use it with Claude max subscription or do I need Claude API?

wow, people really like Pi. and with good reason, it has very good fundamentals. i’ll probably drop the ACP middleman and use the RPC directly for a deeper integration

@plainlogic24 @picoito you can use OpenRouter inference via OpenCode or Pi which Paseo supports

@moboudra @picoito Paseo supports openrouter ?

It's Saturday Had to go do stuff Should be at home Grinding on codex Who else?

@serbedmot @theplgeek glad to hear it, feedback is welcome!

@theplgeek @moboudra thanks for sharing. this is EXACTLY what i have been looking for. incredible work @moboudra

Think I've finally found the one (for now): - open source - Full x-platform support (linux yay) - web client - mobile apps - remote daemon - Codex, Claude, OpenCode support Thank you @moboudra github.com/getpaseo/paseo