Name: Kumiho Memory
Rating: 5.0 (2 reviews)

Kumiho is a belief-based cognitive memory layer for teams running agents across functions. It preserves shared context, decisions, code_why, and how beliefs change—so every agent can build on the team's reasoning instead of starting from isolated chat history.

Hi Product Hunt! Kumiho starts from a different premise: AI memory should hold beliefs that evolve, not just facts to retrieve.

When evidence changes, memory should preserve the earlier belief, record why it changed, and connect decisions that may need a second look. That is the foundation of Kumiho’s belief-based cognitive memory system.

• Decision Memory keeps choices with alternatives, rationale, and evidence.

• code_why traces a code change back to the decision behind it.

• Decomposition links facts and relationships so context stays traceable over time.

We formalize this architecture in our arXiv paper, “Graph-Native Cognitive Memory for AI Agents: Formal Belief Revision Semantics for Versioned Memory Architectures.”

https://arxiv.org/abs/2603.17244

The memory library and plugins are open source; Kumiho Cloud is the managed service. Watch our 47-second demo, explore the plugins, or get started:

https://kumiho.io

https://github.com/KumihoIO/kumi...

What should your AI remember about how a decision changed?

Hi Product Hunt! We're the team behind Kumiho Memory, launching here in a few hours.

We built Kumiho around a problem we kept running into: an agent might remember what we decided, but lose why we decided it, which alternatives we rejected, or what changed later.

Hi Product Hunt! We're the team behind Kumiho Memory, launching here in a few hours.

We built Kumiho around a problem we kept running into: an agent might remember what we decided, but lose why we decided it, which alternatives we rejected, or what changed later.

What I appreciate most about Kumiho is that I can use its plugins directly in both Codex and Claude. It fits into the tools I already use instead of asking me to move to another workspace. This makes it much easier to carry context between agents, continue long-running projects, and avoid repeating the same background every time. I also like that it preserves the reasoning behind decisions, not just the final answers.

One improvement I’d really like to see is support for using asset-related skills that are currently available locally on the web as well. Being able to access the same skills and workflows from a browser would make it much easier to continue working seamlessly across different environments.

I considered relying on each tool’s built-in conversation history or keeping notes in separate documents. Kumiho felt more convenient because it connects context across Codex and Claude instead of leaving it isolated in one conversation or one tool.

Hi Product Hunt! Kumiho starts from a different premise: AI memory should hold beliefs that evolve, not just facts to retrieve.

• Decision Memory keeps choices with alternatives, rationale, and evidence.

• code_why traces a code change back to the decision behind it.

• Decomposition links facts and relationships so context stays traceable over time.

We formalize this architecture in our arXiv paper, “Graph-Native Cognitive Memory for AI Agents: Formal Belief Revision Semantics for Versioned Memory Architectures.”

https://arxiv.org/abs/2603.17244

The memory library and plugins are open source; Kumiho Cloud is the managed service. Watch our 47-second demo, explore the plugins, or get started:

https://kumiho.io

https://github.com/KumihoIO/kumi...

What should your AI remember about how a decision changed?

Kumiho Memory

Shared memory for agent teams, with the why intact

Shared memory for agent teams, with the why intact

Forum Threads

What should AI agents remember beyond facts? We're building Kumiho Memory.

Forum Threads

What should AI agents remember beyond facts? We're building Kumiho Memory.

What's great

What needs improvement

vs Alternatives

What's great

What needs improvement

vs Alternatives