Spoiler: it's not actually that easy. Compaction, security, sandboxing, planning, custom tools--all this is really hard to get right.
We're about to launch an SDK that gives devs all these building blocks, specifically oriented around software agents. Would love feedback if anyone wants to look: https://github.com/OpenHands/software-agent-sdk
How autonomous/controllable are the agents with this SDK?
When I build an agent my standard is Cursor, which updates the UI at every reportable step of the way, and gives you a ton of control opportunities, which I find creates a lot of confidence.
Is this level of detail and control possible with the OpenHands SDK? I’m asking because the last SDK that was simple to get into lacked that kind of control.
If you're looking for open source agents, which can run locally, in Docker, or in the cloud, and which have a consistent track record of acing benchmark scores like SWE-bench, check out https://github.com/All-Hands-AI/OpenHands
We're about to release our Agent SDK (https://github.com/All-Hands-AI/agent-sdk/) which provides devs with all the nuts and bolts you need to define custom prompts, tools, security profiles, and multi-agent interfaces
I’d like to see a “no jerks” license. It’d be MIT by default, but call out specific bad actors as being disallowed from using the software. That way your average corporate user wouldn’t need to consult a lawyer before adopting
Presumably the license would, like practically all open source licenses, be irrevocable. You aren't guaranteed new versions will be issued under the same license (short of a contract saying otherwise, just like every other piece of open source software) but the existing license that did not list you as a jerk can't be revoked...
True, but that's still a risk that adds to the risk of the authors switching the license.
BTW, if the jerk list is tied to the license, if the project had external contributors, they all need to agree to add or remove someone from the list, like any license change…
> BTW, if the jerk list is tied to the license, if the project had external contributors, they all need to agree to add [...] someone from the list, like any license change…
Not if you base this off a license like MIT that allows sublicensing under more restrictive terms (not a lawyer, not legal advice)
This software shall not be used for evil. With the exception of IBM, who, together with their partners and minions, are allowed to use this software for evil.
If you're interested in running agents (specifically for software dev) inside a sandbox, OpenHands [1] runs in Docker by default, and can run on Kubernetes or on a raw VM. It gets access to the standard tools (e.g. file ops, bash) as well as a web browser and a Jupyter notebook.
I have nothing but good things to say about OpenHands . I’ve been using it daily for months now - some local, some CLI, some cloud-based - and while it’s not quite as polished as Devin, it’s as capable and is growing fast.
All Hands is incredibly friendly and responsive to feedback as well, and that means a lot.
Not having all the time in the world to learn, apply, and compare, each one of the many agentic offerings, it would be great if OpenHands would explain what distinguishes it, apart from being the "leading" option... What are the design choices, trade-offs and benchmark results?
I must say it's great that AllHands releases specifically trained models for their tools-use, but then: how does that compare with the Harmony stuff eg with gpt-oss?
yes that's the benevolent take,but in a market for attention, it feels like a bit of a vampire attack to a brand to spun off OpenBrand.
seems some what of a template now, OpenCut was recently on HN because some chinese company already owns their trademark, which is a pun/open version of CapCut, so things are getting a bit hairy at least, lol
I still sometimes have to refer to OpenHands as "formerly known as OpenDevin". It may be limiting in the long run, but it definitely had some short-term punchiness.
reply