Opens beta for Kemory: one memory layer across every AI one uses providing a way past the cold start for India’s own AI model builders.
NEW DELHI, Oct. 1, 2026 /PRNewswire/ — SeKondBrain opened public beta access to Kemory in India, with Indian users’ context held on Indian soil from day one with servers in India making it ‘For India From India’. Kemory, is a permissioned memory layer that remembers what a person has told any AI, so they don’t have to re-explain themselves every time they switch between them. Sitting independently, Kemory operates outside any single AI model, so that the knowledge a person establishes with AIs and agents is available to every other AI they use. The product has been built on the premise that AI’s real limitation today is not intelligence, but memory, and how often AI suffers from amnesia and hallucination.
Indian users are among the most multi-AI in the world, moving between assistants by price, availability and task, using one LLM today, and another tomorrow with an Indian model next month, each one great at something. Every switch means starting over: re-explaining context, re-uploading material, using more tokens, rebuilding what one already knows and wasting time spent correcting the output once again. Push any model past what it can hold, and it fills the gap with a plausible answer built on nothing but assumptions. That cost compounds daily for a knowledge workforce moving faster between AI tools than almost anywhere else.
A Brain Is Grown, Not Built Overnight
SeKondBrain’s belief is that this isn’t a shortage of AI model capability, models have never been sharper. What’s been missing until now is an independent place for knowledge and meaning to accumulate: a layer that holds understanding the way a map holds a landscape, so knowledge, token use and time compounds instead of resetting with every new chat, project or model that comes along.
That belief shapes how the company is building. Rather than launching a single, fully-formed platform, SeKondBrain is releasing its architecture in layers, with memory as the foundation everything else stands upon: first Kemory, the shared independent and permissioned memory layer; then the reasoning and product-building tools built on top of it; followed by the same principle applied to whole organisations, through a business edition, KompanyBrain, already in beta with design partners across the legal sector; and, in time, the flexibility to hold the way any individual works and lives.
An Open Invitation to India’s Model Builders
The cost of moving or recalling isn’t only paid by users. It’s the reason a new model struggles to win users in the first place. India is backing a generation of home-grown AI through the IndiaAI Mission, but every new model still sees its users as a stranger, no history, no preferences, again starting at zero regardless of how capable it is. The incumbent’s advantage is rarely capability; it is the accumulated knowledge and context a user cannot easily take elsewhere with them. That is the cold start, and it is the largest single limiter for any challenger model. Kemory removes this barrier: a user who holds their chats, memories, agents and context in Kemory arrives at a new model already understood, from their first project. The model does not have to earn the history back. It only has to be better.
“India is not a follow-on market for Kemory. It is the market,” said Sachin Dev Duggal, Founder and CEO of SeKondBrain. “The thing standing between an Indian model and a user is almost never quality, it’s that the user’s valuable knowledge and context lives somewhere else. That’s a poor reason for anyone to lose. If a person carries their own understanding with them, any model can be useful to them from the first chat, and the question becomes which one is actually better – that is the question that ought to decide it. This is an open invitation to everyone building models and using AI in India: your users and models start with the knowledge they already own.”
This is not a route to anyone’s data. Kemory holds knowledge for that person, their project or their agent; a model provider is never handed it, it sits independently and permissioned. The user grants access per connection, can see precisely what is shared, and can withdraw it at any time. SeKondBrain is also open to integration conversations with any Indian model provider.
Held in India, and Held for You
Kemory is a permissioned system that any AI, agent or tool can read from and write to, so a person’s knowledge travels with them rather than resetting with every new chat, model or provider, resident in India, not mirrored, for Indian users. Understanding compounds instead of being rebuilt from scratch: what is explained once stays known, and the token cost of that knowledge shrinks over time rather than growing with it. In the company’s internal testing on LongMemEval-500, a public benchmark for long-horizon memory recall, Kemory scored 89.6%.
A New Category, Not Another Assistant
SeKondBrain positions Kemory as infrastructure rather than another application competing for attention alongside the AI tools people already use. It’s built model-agnostic by design: the memory and knowledge a person builds inside it works identically whether they are using a global model or an Indian one, and no provider is given preferential access or integration advantage over another. That neutrality matters most precisely where switching between models is most common or needed depending on the model’s skill.
A memory layer tied to one vendor’s incentives functions as a retention mechanism; a memory layer that can work everywhere is only useful if it belongs to the person using it, not the company hosting it. That is the basis of what SeKondBrain calls knowledge ownership: understanding built through AI is treated as an asset the user holds outright, inspects, exports, shares or deletes on their own terms, wherever knowledge is held.
What’s Included in the Free Edition of Kemory
- Kora for Chrome – import the existing AI history
- Two connected AIs
- 2,000 memories and chats
- Fair-use credits
Referrals extend the plan: every person a beta user brings in who signs up and adds memories earns both accounts 1,000 further memories, permanently.
What Comes Next
Free is the starting point, not the ceiling. Upcoming plans for a Community Edition will release the memory layer under an Apache 2.0 open-source licence, with self-hosting genuinely supported rather than offered as a footnote. A Business Edition will extend Kemory to organisations that need permissioning, knowledge retention, trackability and cost effectiveness at company scale.
SeKondBrain’s KompanyBrain, follows this path and is already live with early design partners doing real client-facing work in the legal sector with more sectors to come. Looking further out, SeKondBrain plans an individual edition that adapts to how any one person operates.
Kemory’s free edition beta is live now at https://sekondbrain.ai/products/kemory with access opening in staggered groups starting with the early sign-ups, by what has been affectionately called the #SeKondBrainTribe.
About SeKondBrain
SeKondBrain is the shared intelligence layer for AIs, humans, and knowledge teams. A neurosymbolic memory and reasoning layer, it is a platform that sits above every model, and below every agent and tool a person or organisation uses, so understanding compounds instead of resetting with each conversation. Kemory, its shared memory layer, is the first layer to be released in a brain being built in layers, memory first. The company’s mission is to unshackle the world’s thinkers from information overload.
SeKondBrain is organised as a flat team it calls Kognitivo: no specialist silos and no pure managers, with everyone building and everyone owning their piece end to end. The company is headquartered in London, with teams in Delhi and Mumbai among other locations. Understanding, unshackled.
For more information, visit https://sekondbrain.ai and https://sekondbrain.ai/products/kemory
For access to SeKondBrain’s Kemory Press Kit click here
Notes to Editors
- Benchmark: cite as “89.6% on LongMemEval-500, internal testing.” Do not round or restate. Full method and results: https://sekondbrain.ai/research
- “Resident in India” may only be used once infrastructure go-live is confirmed.
- Community Edition is not yet released. Apache 2.0 and self-hosting are forward-looking and must remain future tense.
- Business Edition: confirm current status before distribution – remains as described here as live with early design partners; verify this remains accurate at time of release.
- “200+ sign-ups” is a beta sign-up figure, not an active-user or retention figure. Confirm the exact number before distribution.
- No Indian model provider is named and no live integration is claimed. The invitation is a standing offer, not an announced partnership. SeKondBrain is never handed a user’s data; access is granted per connection, by the user, and can be withdrawn at any time.
- This release makes no compliance or regulatory assertion. Any reference to Aadhaar, DigiLocker, UPI, DEPA, the DPDP Act or Bhashini would be contextual only, not a claim of participation, endorsement, integration or compliance, and none is made in this version.
- “Model-agnostic” refers to Kemory’s architecture working identically across model providers; it is not a claim of formal integration or partnership with any named provider unless separately announced.
- “Knowledge ownership” describes user-controlled export, inspection and deletion of memory data. It is a product/design claim, not a legal or regulatory characterisation.
- Confirm whether the legal entity name is required on Indian materials.
View original content to download multimedia:https://www.prnewswire.com/in/news-releases/sekondbrains-kemory-launches-with-indian-users-data-held-on-indian-soil-302895916.html
