Hi everyone! To everyone using, sharing, and contributing to Hive—thank you so much for helping this project evolve.
I wanted to give you an update on the current state of Hive. You might have noticed a few active discussions and feature requests on the tracker recently:
As the creator, I am reading all of these. I already have ideas and designs in mind for them, but if it seems like things are moving slowly right now, that is intentional. I want to take my time to think things through and design them properly.
In the world of agent harnesses, a lot of ideas sound amazing in theory, but in practice, they don't add much benefit and often introduce new problems. Right now, Hive is in a sweet spot: it is simple, practical, and well-designed. While there is definitely room for improvement and new features, I want the project to stay focused and avoid the FOMO/hype cycle.
The Roadmap: 3 Big Features
Moving forward, I am planning three major updates:
- Hive Network: This will focus on improving memory and leveling up the UI/UX.
- Compacting Logic Improvements: This will make long-running sessions much more stable.
- Integrations & Harnesses: (More on this below).
Thoughts on Integrations: OpenCode, Clawd Code, and Copilot
I want to share a bit of backstory here. Originally, my plan was actually to build Hive as a Claude Code plugin. I had to pivot away from that because Claude Code wasn't open source—it didn't give us the ability to define first-class tools, custom prompts, or deep hooks. Because of that, OpenCode was the clear winner for our harness.
But things have changed. As I’m currently on vacation, I saw the massive news about the Claude Code source leak and how the community has quickly rallied to build the open-source alternative, "Clawd Code."
Because we now have access to that underlying architecture, building first-class Hive support for Claude Code isn't just possible again—it's going to be incredibly robust.
That being said, we are absolutely not dropping OpenCode. OpenCode has a phenomenal ecosystem and will continue to evolve. Instead of picking a side, Hive is going to integrate with both and take the best practical aspects from each side.
To keep our focus tight, however, I am planning to drop first-class support for the GitHub Copilot Chat harness. I prototyped some plugins for it, but honestly, it’s just not good enough to support the workflows we want. OpenCode (and soon Clawd Code) are miles ahead.
The Hive Philosophy
This all ties back to what Hive is supposed to be. Whether we are running on OpenCode or Clawd Code, the goal is to be the best practical tool for developers.
My focus is strictly on solid architecture and a developer-first mindset. We are steering entirely clear of the "agentic autonomous nonsense" where agents just spin in endless loops doing whatever they want.
Hive is about structured planning, human oversight, and getting real work done. It stays practical, and it stays production-ready.
Next Steps
My plan is to study Clawd Code, related projects like [openclaw-auto-dream](https://github.com/LeoYeAI/openclaw-auto-dream), and the great suggestions you've left in the issues. I want to cherry-pick the most practical methods from these projects and integrate them directly into the Hive plugin.
Thanks for your patience and for building alongside me. More updates to come!
CC @imarshallwidjaja
Hi everyone! To everyone using, sharing, and contributing to Hive—thank you so much for helping this project evolve.
I wanted to give you an update on the current state of Hive. You might have noticed a few active discussions and feature requests on the tracker recently:
As the creator, I am reading all of these. I already have ideas and designs in mind for them, but if it seems like things are moving slowly right now, that is intentional. I want to take my time to think things through and design them properly.
In the world of agent harnesses, a lot of ideas sound amazing in theory, but in practice, they don't add much benefit and often introduce new problems. Right now, Hive is in a sweet spot: it is simple, practical, and well-designed. While there is definitely room for improvement and new features, I want the project to stay focused and avoid the FOMO/hype cycle.
The Roadmap: 3 Big Features
Moving forward, I am planning three major updates:
Thoughts on Integrations: OpenCode, Clawd Code, and Copilot
I want to share a bit of backstory here. Originally, my plan was actually to build Hive as a Claude Code plugin. I had to pivot away from that because Claude Code wasn't open source—it didn't give us the ability to define first-class tools, custom prompts, or deep hooks. Because of that, OpenCode was the clear winner for our harness.
But things have changed. As I’m currently on vacation, I saw the massive news about the Claude Code source leak and how the community has quickly rallied to build the open-source alternative, "Clawd Code."
Because we now have access to that underlying architecture, building first-class Hive support for Claude Code isn't just possible again—it's going to be incredibly robust.
That being said, we are absolutely not dropping OpenCode. OpenCode has a phenomenal ecosystem and will continue to evolve. Instead of picking a side, Hive is going to integrate with both and take the best practical aspects from each side.
To keep our focus tight, however, I am planning to drop first-class support for the GitHub Copilot Chat harness. I prototyped some plugins for it, but honestly, it’s just not good enough to support the workflows we want. OpenCode (and soon Clawd Code) are miles ahead.
The Hive Philosophy
This all ties back to what Hive is supposed to be. Whether we are running on OpenCode or Clawd Code, the goal is to be the best practical tool for developers.
My focus is strictly on solid architecture and a developer-first mindset. We are steering entirely clear of the "agentic autonomous nonsense" where agents just spin in endless loops doing whatever they want.
Hive is about structured planning, human oversight, and getting real work done. It stays practical, and it stays production-ready.
Next Steps
My plan is to study Clawd Code, related projects like [openclaw-auto-dream](https://github.com/LeoYeAI/openclaw-auto-dream), and the great suggestions you've left in the issues. I want to cherry-pick the most practical methods from these projects and integrate them directly into the Hive plugin.
Thanks for your patience and for building alongside me. More updates to come!
CC @imarshallwidjaja