Build cloud agent coworkers that take on real, multi-step work across the tools you already use. Describe the outcome you want, and Skydive creates a working agent in minutes. No code, no prompt engineering, no workflow wiring required. Skydive agents live in your stack, execute repeatable work, and get sharper over time.
@dhruvtruth It really is incredible how much we use Skydive internally
@dhruvtruth @marcuslowe_ @zariazinn fave thing ive ever worked on tbh / so proud to be a part of it all + work alongside absolute weapons (+ formidable leaders on the CC!)
Congrats on the launch guys! Seems like a pretty solid productivity tool.
@tasbin thank you Tasbin! + yes can confirm Skydive has supercharged my productivity by a million. best part still is that productivity has become entry point - the elevated outcome in conjunction is what's golden
This feels especially useful for repetitive work that usually gets stuck between different tools. How much control do teams have over what an agent can access and change?
@amybradley A lot! You choose exactly which tools each agent can access and can scope permissions down or up to whatever you would like. You can also separately control who on your team can use or edit the agent.
@amybradley perfect question Amy, truly its full control - you decide what it can access and when, set the schedule, set the output, train and when needed, grow its responsibilities.
@amybradley permissions are so important! so we provide options to make your agents private to yourself or your team
@amybradley Hey Amy! control is pretty granular - which tools + data each agent can touch, scoped up or down whenever, visibility set to private / internal / external (for example, personal-use agent vs team-wide-access agent). The most impressive piece for me has been agent posture around sensitive stuff + their ability to flag exposure risk before I've even thought to :)
the self improving part is the bit id want to poke at. learning from previous conversations only works if those conversations carry a correction signal, and most of mine dont. the agent does something slightly wrong, nobody says anything because it was close enough to ship, and now that sits in the history as an example of getting it right. so the primitive i keep wanting is a way to mark a past interaction as wrong after the fact, ideally weeks later when i finally notice. without that the learning loop just hardens whatever it happened to do first. separately, if every agent has its own machine, the thing id want to know before hiring one is whether that machine has a snapshot i can roll back to.
huge congrats to dhruv and the whole skydive team on the launch. been following this since i signed up - agents that actually do the work across your tools instead of just chatting about it is exactly where this needs to go. pumped to dig in. rooting for you guys
@taylor_matthew_desseyn Thanks so much Taylor! Appreciate the support.
@taylor_matthew_desseyn Thanks for the support Taylor!!
@taylor_matthew_desseyn Thank you Taylor! Appreciate it!!
@taylor_matthew_desseyn Thank you Taylor!! 'Agents that do the work, not just talk about it' might have to be our next marketing campaign line?!
Running cloud agents that actually hook into your existing tools without a hassle fixes a huge orchestration pain point. Congrats on shipping!
@thisiskp_ Thanks KP! Really appreciate it.
@thisiskp_ for real! plug in all your favorite tools -> maximize your agent capabilities
@thisiskp_ thank you KP! orchestration overhead was the exact tax we wanted gone. describe the outcome, skip the wiring, boom
The real test for me would be whether these agents still have time after the initial excitement wears off. if they can reliably take repetitive work off the team every week, that could be really valuable.