惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

罗磊的独立博客
SecWiki News
SecWiki News
酷 壳 – CoolShell
酷 壳 – CoolShell
爱范儿
爱范儿
量子位
M
MIT News - Artificial intelligence
GbyAI
GbyAI
cs.AI updates on arXiv.org
cs.AI updates on arXiv.org
K
KPMG report finds enterprise disconnect between AI and its ROI | CIO
TaoSecurity Blog
TaoSecurity Blog
博客园 - 【当耐特】
H
Heimdal Security Blog
腾讯CDC
The Last Watchdog
The Last Watchdog
Security Archives - TechRepublic
Security Archives - TechRepublic
Hacker News: Ask HN
Hacker News: Ask HN
S
Schneier on Security
Microsoft Security Blog
Microsoft Security Blog
WordPress大学
WordPress大学
博客园 - 司徒正美
Recent Commits to openclaw:main
Recent Commits to openclaw:main
C
Cybersecurity and Infrastructure Security Agency CISA
S
SegmentFault 最新的问题
大猫的无限游戏
大猫的无限游戏
Application and Cybersecurity Blog
Application and Cybersecurity Blog
F
Full Disclosure
有赞技术团队
有赞技术团队
T
Tailwind CSS Blog
Engineering at Meta
Engineering at Meta
cs.CV updates on arXiv.org
cs.CV updates on arXiv.org
T
Threatpost
月光博客
月光博客
A
Arctic Wolf
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
雷峰网
雷峰网
T
Troy Hunt's Blog
freeCodeCamp Programming Tutorials: Python, JavaScript, Git & More
The Cloudflare Blog
D
DataBreaches.Net
O
OpenAI News
L
LINUX DO - 最新话题
宝玉的分享
宝玉的分享
小众软件
小众软件
V
Vulnerabilities – Threatpost
A
About on SuperTechFans
人人都是产品经理
人人都是产品经理
T
The Exploit Database - CXSecurity.com
Martin Fowler
Martin Fowler
美团技术团队
P
Privacy International News Feed

HN's home page

Rainbow Query Language | Hacker News Exec into Node via Kubectl An AI native hedge fund The Seven-Action Documentation Model | Hacker News Package Manager for Kubectl Plugins Tongan Castaways | Hacker News Tech overlords plan for conscious AI to conquer the cosmos. What could go wrong? Data Breach Disclosure Lag Is Getting Worse How LLMs Work | Hacker News I Dropped PRDs for Shape Up Go Experiments Explained | Hacker News FCA's Palantir deal could expose UK financial data to Trump's US, critics fear WebXR BCI for Neural-Adaptive Avatar Control in Mixed Reality The first murder conviction via DNA analysis Tom Interviews Theo de Raadt of the OpenBSD Project (2019) [video] Show HN: Replace shell commands with bun shell typescript scripts Quay.io Is Down | Hacker News AI driven analysis of brokerage account fees in the UK Bill Gates Spent Years Crafting His Image. Now It's Cracking Using LLMs to secure source code Wi-Fi 8 in the Lab [video] The household battery revolution that could change energy bills and the world Is Python Becoming Pinyin? | Hacker News Livia – Executive Assistant | Hacker News FindMyPipe – Query Apple Find My from Linux for AI Agents Show HN: Agent skill for creating product launch videos with Remotion RecruitMyself – AI job search copilot for resumes and applications AI coding agents and the erosion of system understanding The 'Resting' Generation and South Korea's Youth Recession AMD Computex 2026: 10 Years of AM4, AM5 Support Through 2029 Docker Networking Explained | Hacker News Textbooks in Tokenland | Hacker News Key Chemistry Question Answered, No Quantum Computer Required Gifts For Retrocomputing Fans – remix yesterday's tech with a modern spin Miscellany № 49: introducing the quasiquote – Shady Characters Amazon Thinks the Future of Data Centers Is a Technical Problem It Just Solved A brief history of the UUID (2017) Flying High Unpressurized (2016) | Hacker News Five Years of Trying to Add Recursion to Lychee How British comfort food won over the French Blorp Language | Hacker News Decache – you might have the internet's lost media in your PC's cache folders Criminal Activities and Migration | Hacker News A free, open-source library of DESIGN.md files for AI-generated UIs MiniMax M3 | Hacker News People are apparently farming citations on ResearchGate – Chuniversiteit Hacker News Basketeer – a typed TS SDK for your Tesco account, with nutrition data 'Penguin' decays from CERN's Large Hadron Collider experiment hint new physics Emergence World: A Laboratory for Evaluating Long-Horizon Agent Autonomy Homebrew lead Mike McQuaid: Sandboxes and Worktrees - My Secure Agentic AI Setup Lean, Not Backpressure | Hacker News AI Dangers Eclipse Nuclear Weapons at Singapore Defense Forum Open source analytics that answers backbase How turkey hacked the hair-transplant industry How GPT Image 2 Is Transforming Marketing Workflows in 2026 Improve Git monorepo performance with a file system monitor Strava for Claude Code MiniMax M3 on Qubrid AI There's Something Else We Should Be Worrying About Celebrity Profile of an A.I. Actress What Is Windows K2? | Hacker News AI is devoid of meaning and humanity. Its vapid voice suits the political moment Show HN: Interpreto – Live Translation for Travel Taxicab Geometry Sealed classes and interfaces in Java (2025) Show HNs | Hacker News My AI Skill Edited This Video That Explains My AI Skill – Arcturus Labs Amazon Pinpoint End of Support The Mystery of the Backward Index MP/M's Process Dispatcher SlimTide Reviews: A Modern Solution for Metabolism and Energy Learning Lustre: Type-safe front end development with gleam Thomas Mann: Goethe Heartened by Panama (As Suez for English, or Danube-Rhine) How to make Message Log of the Unreal Engine 100 times faster Sum-product, unit distances, and number fields Can Meta Buy Belief? | Hacker News Twenty Years of Bigtable | Hacker News Show HN: Combine WigglyPaint GIFs into Video Show HN: AgentThreatBench – Benchmark for AI Agent Memory Security Genius Spotted in the Wild Napkins: Where Ethernet, Compaq and Facebook’s cool data center got their starts (2011) Moderate caffein use alters sleep-related EEG Nvidia Announces RTX Spark | Hacker News Show HN: Ministry of Everything – CLI agent harness for a single operator CEOs blame AI for layoffs, MIT prof says it fits a pattern to find cover story Bugs I didn't expect while building a zsh cleanup script for macOS dev machines Nvidia jumps into PCs with new chip debuting in laptops from Microsoft, Dell, HP Nvidia unveils PC 'superchip' in challenge to Apple and Intel Show HN: Having fun making mini static site apps Synthea API: Create Synthetic Medical Records as a Service Berkshire Hathaway to buy Taylor Morrison for $6.8B in cash The most complex model we understand [video] SanDisk stock is +4,440.53% in the past year Driftwm: What if your window manager worked like a whiteboard? US Immigration enforcement looks into buying ad data AI Is Creating More Work for Australia's Workplace Tribunal Finding New Biblical Cross-References with Codex Glide: A tiling window manager for macOS Ultra-highly efficient enrichment of uranium from seawater via studtite nanodots (2024)
Show HN: Paseo – Beautiful open-source coding agent interface
timhigins · 2026-06-03 · via HN's home page
Show HN: Paseo – Beautiful open-source coding agent interface (github.com/getpaseo)
90 points by timhigins 4 days ago | hide | past | favorite | 52 comments
 help


I'm the maintainer of Paseo. I didn't submit this, so it was a nice surprise to see it on HN!

I'm around if anyone has questions about the project.


Any idea if claude code will continue working with paseo after the billing/usage changes in a couple weeks?


Claude Code (via the subscription) will continue working under Paseo but it will consume a different pool of credits, which depending on your sub you get different amounts. Practically speaking you will be able to use only a fraction of your usage in Paseo, this applies to any programmatic usage of Claude Code.

This might help: https://support.claude.com/en/articles/15036540-use-the-clau...

Paseo has a terminal if you want to keep using the TUI inside of it, works on mobile and desktop.

I will keep supporting the Claude Code harness in Paseo, as many users are using it via direct API usage and sometimes even with Chinese models.


Gotcha, I guess I should have asked if you were using 'claude -p' or something else to re-use the subscription.

It looks pretty interesting though, I'll give it a shot


very nice and well executed product for what i assume is a small team! (of one?)

i guess the obvious question is... if its FOSS and local only, whats the business game plan? you've impressed me that you can ship, but you've basically just matched what eg devin desktop and cursor now do. what next?


Thanks. Yes, team of one right now.

I think there's still a lot to do and it feels like I'm only scratching the surface of what's possible in the local layer, so matching what Cursor does is not the end game for me.

I want to keep the core open and local-first, the business, when it comes, will be around the convenience/team/enterprise layer.

Right now I am letting real usage guide me by paying attention to what teams using Paseo need.


good luck sir. if you ever want a job i'm happy to refer you to cognition/devin.


You can already use Gemini CLI with the ACP adapter, go to the provider settings to enable it.

Antigravity does not yet support ACP, when they do, it will be added to Paseo


"Ship on the go" is so insane to me.

It's like people pulling their phones out while taking a piss standing, or having to pull their phones out when the traffic lights are red in a crossing.

Just, do one thing at a time, live a life.

Shipping code from your phone, whhhyy. Mates, this isn't a flex, it's depression.


Maintainer here. I get the concern, but that has not been my experience.

I don't think of the mobile side as a way to keep working on your phone on top of your full working day, although I will admit it can be used that way.

It is more about being able to step away from the desk without losing access to the work. Let's say you spend 6 hours at your desk, what if you could spend 2 of those hours walking instead?

To me, it has been valuable for steering long running agents, brainstorming ideas or triaging PRs/issues while taking a walk.

Fun fact, Paseo means "stroll" in Spanish, which is where the name came from.


I do sometimes wish I can monitor my agent on the go and keep drip-feeding it tasks. Because I do genuinely need to read how the last task went before I decide whether to correct or to move to the next task.


Cool project. I don’t think people will get the mobile version until they need it but when they do it’s a mind bending, life changing realization. I built my own IDE to have it on my phone because I have small kids and it is truly life changing. I get to spend way more time with my kids while still getting work done.

If you’re open to collaborating I’d be happy to share what I’ve learned an see if we can share resources or lessons.

https://calendly.com/ryanwmartin/open-office-hours


This seems extreme. Maybe I’m just optimistic but I think people can be intentional and present while also having the convenience and accessibility that something like paseo offers.

- If I’m on-call, it would be great to just take my iPad mini around with me instead of my heavy MacBook Pro. - Sometimes I’m on the couch with my phone and want to query an agent with access to my computer’s file system.


I absolutely get that, but I've been working with Claude remote control on my phone as I bicycle every other day. (I have heart and mobility problems and I have to take it gentle with exercise.) I ride to a beautiful spot where there's a bench and sit and use Claude to write features or React components. I'm actually wanting to try this to see if I can rig up a voice-control flow so I can literally just ride my bike all day from park to café to wherever and still be getting code done.

That's the dream, right?


It's very addictive when you're working on something cool and the agents are iterating nicely. Instead of browsing reddit / HN / instagram etc during downtime, I find it much more fun to build something.


Umm. I have small kids that I take to classes and swim school and gymnastics and so on and so on. I built www.propelcode.app which is similar to this project so I didn’t need to be chained to my desktop and I didn’t need to let AI agents run wild with no code review or human in the loop oversight. For some of us doing development on our phones is living life.

And the fact that some of the coding tasks I trigger take 20 minutes means I can fire off a message, leave it, look at the code when I have a spare moment, suggest changes, and go on spending time with my kids.

Honestly it’s freaking amazing.

And I will definitely check out this project and contribute to it with what I have learned! Kudos to the developers behind it.


Have been using it for a few days. It's chuck full of features and works great. (Have not used Relay, direct only)


Not to undercut the open source nature of this, but what makes this "beautiful"? From a design standpoint, it's basic tailwind. Neutral grey tailwind at that, using Lucide icons. There's nothing wrong with these, but it'd be more apt to say that the design is unopinionated. It's the default choice when design intent is the afterthought and a focus is on functionality.

Again, not trying to undercut - looks like a solid agent interface, it just struck me as strange that beautiful was the adjective chosen when design seems to not be the objective here.


If you were to compare it to a painting or to the Grand Canyon or to the Northern Lights or like an act of kindness or a parent's love for their child or something, then I guess fine, not beautiful.

But for an open source project it's very nice!

(Note: none of the marketing materials for the website chose that word, at first glance. It seems to just be a descriptor given by the HN poster.)


I'm the maintainer of Paseo. This is correct, I do not use the word beautiful anywhere.

I personally do think it's beautiful (obviously), but I would not use that word in marketing materials, I'd rather people judge from seeing the screenshots or trying the product.


If anything it’s unglued me from my computer. I’ve been able to keep an agent working on a project while on long runs, bike rides, in transit. Much of our development workflow is the human in the loop refinement cycle now.


When I see "Beautiful", I automatically ignore it. Beauty is subjective, fashion changes. Bootstrap or React were considered beautiful in their time.

So basically an open source agentic GUI. Instead of "beautiful", it should emphasize what makes it special. Is it fast or lightweight? Does it do something other tools don't? Or does it do it better? What's it killer feature?


(Designer here) I like checking up what a product looks like when it's pitched as "beautiful". Mostly there are two ways to really meet that promise:

    1. Masterful application of a trend (Stripe, Raycast)
    2. Strong and recognizable personality (!boring, Notion, OG Basecamp)

Option 2 is the most accessible to small teams, but it's not an intuitive conclusion to draw. Both need an experienced designer to succeed, but option 1 sounds like it's a safe bet instead of a leap.

Most claimed "beautiful" products result from work done without the experience & taste to tell option 1 apart from an attempt at option 1.

In reality, you can pull off a strong personality and a clumsy execution, whereas following a trend clumsily looks like failing to read the room, and leaves you looking dated almost instantly.


I do think it's possible for one to qualify/quantify the how and why they deemed their product "beautiful".


As noted above, I'm not the creator of Paseo, just a big fan. Beautiful is just ways to describe it (along with convenient, and powerful) since it has a very focused and clean UI. Especially compared to many open source projects, which often don't put in that much effort or are unabashedly vibe-coded.

Of course it is open source so I hope some of the designers/people who've commented on this post can maybe contribute ideas to improve it even more. I have noticed a few places where some common actions take 1 or 2 more clicks or taps than they should - things can always be a bit more convenient and beautiful.

But overall I'm incredibly impressed and you can see some examples of the focus on simplicity and a nice UI, and follow the creator here: https://x.com/moboudra


Looks nice. The README pitches model-per-task picking but doesn't say much about context management. In coding-agent loops the full system prompt + tool specs re-send on every step — a 30-step task pays the input cost 30x. Prompt-cache headers catch the static prefix, but the per-step diff (file diffs, observation tokens) isn't cached, and that's often most of the input. Auto-summarizing older trajectory into a state vector saved 40-60% input tokens in workloads I've looked at — could be a useful daemon-side concern since users won't reach into each agent's internals.


This looks awesome! I've been working on a fairly similar project but just for myself (definitely not production ready) and this format for interacting with code seems like it has a lot of potential. One thing I was considering as a value-add was that this type of interface enables is having support for MDX output or even embedded MCP UI apps in the middle of a chat. Typical CLI agents don't support that but a UI that lets your code embed Excalidraw diagrams and graphs could be great for things like data analysis on the go. Even embedding images in a chat would make using coding agents a lot nicer imo.


Maintainer here. Great points! Paseo already supports embedding images in the chat, but no MDX or chart rendering yet. I have been meaning to add Mermaid support but haven't got around to it yet.


Nice work. I've been building a token usage dashboard for Claude Code — seeing how much each project actually costs. Would be curious if Paseo surfaces any token/cost visibility too.


Cost and token usage is surfaced per session in the context meter, no per-project summary though.

Working on getting usage (5h/7d) in the app, it has been a very requested feature.


Designer here, this is what finally unlocked multi-agent workflow for me. Each session comes with a script that runs a local server in a different port


Fellow openchamber user here, it's been pretty great. I love competition or more offerings though! I'll be watching Paseo.


I'm using Gitea itself as my coding agent interface. I simply tag @codex or @claude on an issue and ask it to open a pull request. Or ask it to reply back on a comment thread, etc.

The Gitea interface is already a pretty good interface that can be accessed from any browser on any device.


So this is an alternative to using one coding agent with openrouter, changing the models between tasks? I am a neophite in these things, my ai use is more calling apis from scripts right now. Can somebody please explain the pros and cons (beyond to openrouter fees) of each?


This looks really neat!

I've tried Conductor, Superconductor, cmux, and a half-dozen other apps that all give you a similar interface. It'd be great if there was a comparison to at least some of those on Paseo's website.


Maintainer here. From the ones you listed, I only have a comparison page to Conductor (see others in the menu): https://paseo.sh/docs/alternatives/conductor

We're all converging on similar interfaces, but there are differences. The main ones right now are:

- FOSS

- Local-first and self-hostable

- No telemetry or forced login

- macOS, Windows and Linux

- Available as a desktop app, web app, native mobile app and PWA

- Daemon/client architecture, run the daemon anywhere you want and connect with any client

- Support for popular agents like Claude Code, Codex, OpenCode and Pi. Plus native ACP support which allows it to support most other agents

---

Paseo supports terminal multiplexing, like Cmux, but I'd consider them completely different product categories.


Nice work.

I like seeing projects that focus on usability instead of adding endless features.

How long did it take to get from the first prototype to this version?


Maintainer here. Yes, the web app is a static Expo export, so the simplest self-host path is to build packages/app and serve the generated dist/ with any static HTTP server. The daemon itself is published to npm as `@getpaseo/cli`.

Docs for this are currently missing, and I should probably package all of this in a Docker image. I'll do that today!


Looks incredible, I just built what seems to be a shittier version of this using OrbStack + JJ a couple days ago. Will take a look!