Probably a feature? OpenHands was stuck in an edit > read > edit > read loop, and I commented within the running thread, which it listened to Was enough to bias it out of the circular thinking Whoever coded this; genius! No other AI chat user interface I've come across can do this mid-flight 🙂

Now, OpenHands does 100% of my infra IAM research for me 😅 Got an IAM error on GCP? Send a screenshot to OH... and it just works!!! Can't imagine going back to the early days without OH: I'd spend an entire afternoon figuring how to get IAM right 😅

Asked openhands to “show me some love” and it coded up this app for me, actually kinda genuinely feel loved 😂 https://work-2-gpdxfetpawgedbub.prod-runtime.all-hands.dev/

I will have to say I super enjoy collaborating with OpenHands to make my project website (the ability to view the website wo running things locally is such a great feature)

OpenHands is developing an OpenHands #C07KHERRM2S tab in our documentation (it actually serves the doc site and check the page using browser), probably also a success story by itself lolsob

Fixed a problem with BitBucket integration for Openhands by providing OpenHands my BitBucket access token through custom secrets + asking it to browse docs 🔥 https://github.com/All-Hands-AI/OpenHands/pull/9021/commits/e0fcde27609bcbcedd62aab69d00c1be06769603

Completed my first real project today with OH. A nginx server block manager. I have a number of domains l host on a vps nginx server. This tool has four options to list the server blocks, add server blocks and add a custom html page to selected blocks.. Reasonably straight forward but enabled me work through the OH functionality. WIP https://github.com/cbay-au/nginx_serverblock_management

One of the cool things about OpenHands, and especially the Slack Integration, is the ability to empower folks who are on the ‘front lines’ with customers. For example, often times Support and Customer Success teams will field bug reports, doc questions, and other ‘nits’ from customers. They tend to have few options to deal with this, other than file a feedback ticket with product teams and hope it gets prioritized in an upcoming sprint. Instead, with tools like OpenHands and the Slack integration, they can request OpenHands to make fixes proactively and then have someone on the engineering team (like a lead engineer, a merge engineer, or even technical product manager) review the PR and approve it — thus reducing the cycle time for ‘quick wins’ from weeks to just a few hours. Here's how we do that with the OpenHands project 🙂

Browsing in CLI is not implemented? No problem! openhands found its way around! 🎉

Agent running... (Press Ctrl-P to pause)

Wow! Thank you for pushing me to check this. Here's what's actually available now: (2024/2025)

It sounds pretty scary, but OpenHands CLI can now actually help you debug Kubernetes issues by running

kubectl

command 🤯 You have to be there watching & approve its command of course, otherwise it might potentially 💣 your infra -- but this is a huge time saver for me, especially since I'm not really an expert in debugging k8s issues and don't know where to start looking 😅

Poking around for a hackathon, and apparently got a very sandboxed environment to behave and run local inference via WebGPU including model downloads. 👀 The future is now

OH running WebGPU would be absolutely 🏀 baller.

I notice this for the first time in a long time: OH is now able to have 3.500+ events long conversations, AND this time, without losing track until 3400 - 3600 somewhere! The conversation did have some small misses, but they were very particular details from microagents which it misses under normal circumstances as well. It didn’t lose track of what it was doing, and it felt like it understood the project better and better 🙏 (llm condenser set at 100 / 4, Sonnet 4, heavy use of “task” markdown files)

hello, I added os.path.abspath() to the workspace_mount_path configuration in config.toml to use relative paths：https://github.com/All-Hands-AI/OpenHands/pull/9615

Working out some thoughts about how to present this as a tool for the community, but I got a bit triggered by the state of the trajectory visualizer. I’d like Vizz to handle ingestion of any convo chat export format, OpenHands trajectories are the standard right now but I’ll be going after Open WebUI and some online platforms like Bolt that don’t readily export as nicely. I strived to make the UI dense with available information from the session, but without being too cluttered (it is too cluttered at the moment). I’ll drop a link to the repo in Slack when I get to a decent checkpoint.

Ideally it runs entirely locally so it can sit up on Netlify or be self-hosted for quick access, DM me if you’re interested in getting involved as its the first tool project I’m planning to fully release in MIT form. Fun stuff!

https://github.com/All-Hands-AI/OpenHands/pull/9589 - I'm parallelizing OpenHands by having one conversation fix exactly one failing unit test file This prompt helps me make sure that there are no conflicts between multiple simultaneous conversations

Please fix the failing unit tests in {{ file_name }}

When you're done please pull any changes, then push your changes to the existing branch

Guys, I just wanted to share my success story. I have fully forked OpenHands, I've created an algorithmic orchestration system, I'm using 7 different AIs, I've created a CTO and senior-level both a senior software stack model, I've created a UI/UX model, I've created different models for different things. Basically, I've created technical prompt personas and orchestration systems and put that in a whole like algorithm system that's governed by like an AI model that's customized just for orchestration, and that's created me. I've got 100% eval, I've got about 85 on SWE, and I'm busy running live code and then other benchmark. I'll show you, I'll share the link. It does cost, because you need to pay for tokens, but I'm not making a profit. I am open sourcing my forked version of OpenHands. Would love you guys' feedback. I'll probably do this in the next few days.

Guys, I will give a full breakdown report as sometimes I've realized now AI didn't test everything. All the problems. So my 100% eval is actually false and 85 on HW is also false. Just guys hold on for a bit before I give you the real facts and I'll actually give you the real JSON so you can actually verify.

I have started using OpenHands to do “deep research”-style tasks for me, which seems to be pretty useful so far bufo thinking

I mentioned OpenHands in a recent article, greatly enjoying use of it with Qwen-3-Coder on Cerebras.

This weekend it also helped me a lot. My fiber Internet has been down since Friday, and tethering from my phone limits the throughput and bandwidth I can use for heavy development. I was able to do a fair amount of work on a new project using the cloud offering, without burning through my mobile data 😆

Just wanted to share that using GPT-5 Pro to rewrite my prompts on my pipelines for Openhands + GPT-5 has made a ton of difference. if you are running something custom on top of openhands and you have a very prompt heavy solution. be sure to try using the prompting guide + your current prompt and getting gpt-5 pro to rewrite it.

My PRs made with and without help from OpenHands over the past 2 years. (Of course OpenHands made the plot)

Had some fun this morning working with the OpenHands CLI + Figma MCP server! • 📽️ Here's a quick loom video: https://www.loom.com/share/4512e2ce43cf45099bee340b31a74d1c?sid=f8ea8451-18bc-46a2-b2ce-b4f850b42523 • Figma Desktop provides an MCP server that I can connect to locally with the OpenHands CLI -- see Figma docs. • I found it worked best when configuring Figma MCP to use

SHTTP

instead of

SSE

I should post here more, but right now I found myself doing a particularly diverse variety of things with OpenHands: 1. Agent 1: Trying to fix the map tile server for the webarena benchmark by porting the files over to s3 and setting up an initialization script 2. Agent 2: Creating a figure of the transformer in Javascript 3. Agent 3: Creating an animation in Javascript of the difference between LM training and inference 4. Agent 4: Doing research on the size differences of the different llama 3 models, and creating a markdown table with the sizes and expected flops All of these are things that I would’ve had to do manually (or probably not done at all due to lack of time), so it’s pretty great to have agents be able to do them.

OpenHands CLI has become immensely helpful for research stuff too -- even when it comes to cleaning up some vibe-coded function lolsob Basically, I have code in a Jupyter notebook that's very messy. I sent an agent to refactor it into a library so it can be reused in many different places. BUT i need to do a quality check to make sure it has the exact same behavior as before. After manually reviewing the refactored code, I tried to set a debugging script myself to check the difference, but it is just VERY boring to check JSON fields one-by-one to make sure they match & update the refactored code to debug (e.g., add print here and there...) And then OH CLI just nailed it 🐕 (and found a stupid bug that was introduced by me, the human 😅 )

can you read

SOME_NOTEBOOK.ipynb

, we have ran this script to processed a bunch of completions in

/mnt/data/research/data/XXX/downloaded_data/**/processed_completions.json

. i have a new implementation for process completions in

/mnt/data/data-sdk/XXX/trace_segments.py

function

completions_to_trace_segments

. Can you help me write a test script that goes through every

processed_completions.json

and makes sure that after loading completions.json, passing it through completions_to_trace_segments, it will return trace segments that match the existing processed_completions.json?

OpenHands-GPT-5 made it! It implemented reasoning, backend and frontend; it started OpenHands on the Saas, and it got screenshots to prove it. That... appears to be what Gemini 2.5 Pro is like, when we're not looking. 🫠

Saw this awesome success story shared in our Discord 👀

Lolz that was me having fun 🎢 ☝🏻skin tone 2 I didn't think any agent had a hope of dusting off a 7-year-old Zero Knowledge Authentication demo in PHP 7.1 Copilot asked to get involved and bombed out not able to install legacy PHP 🔧 😂 Here is the full PR where it stars out with Copilot having a fail then all the later commits are Openhand driven on an iPhone 16 while waiting for a flight at the airport 🛫 😂 https://github.com/simbo1905/thinbus-php-srp-demo/pull/4