Tanka has officially launched EverMemOS, an operating system that brings persistent memory to AI assistants with the aim of making them as consistent and context-aware as a human teammate.
Most AI tools still start from zero every time a new conversation begins. EverMemOS changes that by creating a memory-native layer that connects your chats, files, and workflows across Slack, Gmail, Notion, and even WhatsApp. That means your AI assistant can now remember what you discussed last week, follow up on meeting notes, or retrieve that one file your team mentioned days ago.
Tanka is already putting this framework into practice through its Fundraising Agent, a workflow-based AI assistant that helps startup founders draft pitch decks, track investor questions, and manage outreach with full historical context. This use case highlights how persistent memory can turn repetitive operational work into an evolving knowledge base, while eliminating redundant inputs and improving response consistency.
Underneath its sleek interface, EverMemOS runs on a structured architecture designed to make AI memory both scalable and searchable.
The system collects data from daily tools, channeling them through a Memory Processing unit that turns chats, files, and interactions into searchable and structured knowledge.
On top of this foundation, the Agentic Layer manages reasoning and workflow execution, while the Memory Retrieval Layer handles context recall through hybrid keyword and vector searches.
This layered design allows EverMemOS to understand continuity — for example, recalling past emails when users plan a follow-up or connecting meeting notes with related documents — bridging the gap between task-specific AI and persistent digital memory.
In early internal tests, EverMemOS achieved a 92.3% LOCOMO score, one of the highest indicators of long-context memory efficiency. This points to a proven capability to retain and retrieve context over extended interactions without lag or data drift.
EverMemOS is built specifically for fast-moving teams and early-stage startups, but its open architecture suggests much broader potential. The company plans to open-source the platform, inviting developers to build custom memory-native agents that extend beyond productivity into creative or enterprise domains.
Founded by a group of former AI researchers and product designers in 2023, Tanka positions itself as a global AI infrastructure company focused on long-term contextual intelligence. With teams operating across multiple regions, the company’s mission is to make intelligent systems capable of learning from accumulated experience — not just reacting to isolated prompts. EverMemOS marks its most ambitious step toward that vision, combining usability and scalability in one framework.
By enabling tools to “remember,” Tanka addresses one of the AI industry’s most persistent frustrations. It’s not just a new model or plugin — it’s an operating system for long-term context, built to make AI genuinely useful in ongoing collaboration.
