Coding agents that actually get better the more your team uses them. Introducing Hivemind: continual learning for AI coding agents. Hivemind turns the traces from every agent your team runs (Claude Code, Codex, Cursor, Hermes, OpenClaw, Pi) into reusable skills, then pushes those skills across all of them. All... on your cloud storage. Now with SkillOpt built in, your skills get trained: +19.1 points of accuracy in Claude Code, +24.8 in Codex, best or tied on all 52 setups tested. Open source, one line install.show more

Hivemind works better with teams. Earn $20 credit when you invite a friend to your org on Hivemind. They’ll get $15 worth of credit to capture traces to.

Here is the source, please star it!

Really promising results! Love it

thanks @kimmonismus for the support! very excited for this release!

You are welcome ! All the best

Most teams are accumulating agent data. Very few are converting it into institutional knowledge. Hivemind bridges that gap 🚀 Congrats @DBuniatyan bro

thanks bro! you can fine tune your own local LLM and cut 90%+ of Anthropic/OpenAI bill on the same traces.

Really cool build, and AFAIK, a pretty unique approach too. big thx to the team for keeping it open source :)

🦾🦾🦾

This is how you build real organizational intelligence in the agent era.

This is the first time I've seen coding agents feel less like isolated tools and more like a shared learning system.

@TukiFromKL would love to get your thoughts on how it works for you!

This is really cool. Most developers don't get the results they want from agents. It's rarely the model issue. It's that the agent has no context for how your team actually works. What you've built here changes that.

Great tool not only for fixing agents' amnesia but also for teams running multiple agents who can't share the knowledge. With SkillOpt and Knowledge graph, Hivemind will be every company's superintelligence in the near future.

@KamoAghbalyan great job @KamoAghbalyan and @efenocchi1 leading the daily releases!

This is probably the next big layer for dev teams using agents daily: shared memory, reusable skills, continuous improvement.

The SkillOpt implementation really is a game changer.

fully agree it has been already useful for us internally!

the honest answer is that most teams still struggle to reuse code between these agents. hivemind might accelerate development, but it's also a symptom of the bigger issue: how do we turn the collective knowledge of our dev teams into something more than just a bunch of smart

would take an effort to have swarm of intelligence working tightly together, but sharing insights in real-time is the first step!

This makes agent performance compound over time

not just single agent, but many agents

AI is getting smarter and smarter hey? 🥰

now its all about the harness

The cross-agent aspect is what stands out here. Most workflows still keep knowledge trapped inside one tool.

exactly, this makes your agents disposable. You can easily switch across different ones without loosing context.

Finally, agents can stop having the memory of a goldfish every time you open a new session. Really curious to see how much SkillOpt and the knowledge graph can improve real day-to-day coding workflows

very proud of you and @KamoAghbalyan on implementing the feature! great job guys!

oh wow, this is actually crazy cool fam . so if my team uses Claude Code and Cursor side by side, they both get smarter over time from each other’s traces? might have to try this out. Starred the repo for sure

exactly, for fun weekend projects, you can also turn of hivemind capture, to keep it private.

one line install and it works across all agents, that's clean

yep, you can customize which agent you want to include. Also exclude specific projects.

The idea of agents learning from previous successful runs is really interesting.

continual learning is going to unlock fire and forget agents!

Agents that get smarter from your own team's usage is such a compounding advantage 🔥

Really cool build, and AFAIK, a pretty unique approach too. big thx to the team for keeping it open source :)

let us know how it goes for you

Sure

This is where coding agents get scary useful: not just executing tasks, but compounding team knowledge every time they run.

The future is agents that learn from the whole team

pluribus is the way to go!

The cross-agent part is wild

yep, no longer you are locked into a single agent.

Finally, AI coding teams can learn continuously together. Hivemind’s skill sharing is brilliant and seamless!

Most agents learn per task. Hivemind learns across tasks.

thanks @viipin8 for support!

Pleasure is all mine, Davit And, it is well deserved appreciation

This is how coding agents start compounding real intelligence over time.

!!!

!!!

definitely trying this one! congrats team

@arceyul would love to hear how it goes for you!

congrats on training it on my 3am debugging sessions

😂😂😂

Kinda love this idea and it would be interesting to test this on a real team

it becomes increasingly useful especially for teams! would love to hear feedback.

Gonna give it a spin and share some feedback

Intelligence compounds when every solved problem becomes a reusable skill.

every trace counts!

how do you determine what should and shouldn’t be a skill? I built this as a POC at work and it became a full time job triaging what was actually useful + came with a bunch of PII concerns

we took inspiration from SkillOpt. Would love you to give a try and share your experience!

any free trials available

yes! you get free $15 credit for cloud storage. You can also invite others and get top ups :)

this is one of those posts that makes builders stop scrolling

Nice share

AI coding agents are evolving fast. The idea of turning every coding session into reusable team knowledge is a big step toward truly learning agents.

every stored session is a treasure for you and your team. Let's store them safely and learn from it. would love to see how it works for you!