Nakshi
Understands you, gets the work done, and collaborates in multiplayer — your AI Team and theirs on one task.
Story
It answers, then forgets you by tomorrow — and the work stays yours.
Nakshi is built the other way around: it understands how you feel, remembers what matters, and finishes the work.
Then a job opens — your AI Team and theirs on one task. Collaborative, multiplayer work.
For you
Nakshi is supposed to understand you, remember what matters, and finish the work. Not another chatbot you have to re-explain.
Same supersystem. Different lives.
Ship without re-explaining the system every Monday.
Nakshi is supposed to be the team around you: hire specialists, keep the decisions, and write the spec in your voice.
Hire a team
eng-lead, backend, frontend — they keep going after you close the tab.
Ink the spec
Draft ADRs and postmortems in Notes, in the way you already write.
Remember the why
Architecture calls land in the Second Brain so the next sprint does not start from zero.
Writing happens in . Memory lives in the Second Brain. Agents finish what you start.
Emotion across the system
Nakshi is built to sense how you feel across the whole supersystem: agents, memory, and workflows, not only in a single reply.
That includes our own emotion-native model, nakshi-anarasi-0.1, and the same emotional context when you run frontier or open-source models on Nakshi.
personal knowledge graph
Long-term memory
We are building a Second Brain that remembers your notes, people, decisions, and context, then connects them so your AI can think with your world, not against a blank slate.
Conversations, voice notes, and work feed one living graph. Agents pull from it when they act, so nothing important has to be re-explained.
Ink AI
Ink is the writing partner in Notes. You set the styles and the tones. Press Space, then Continue, Improve, Translate, or ask it to draft from files and other pages.
What lands on the page feeds the Second Brain. Agents can pick it up from there.
Q3 launch brief
Acme-pricing.pdf
1.2 MB
Nova-annual.xlsx
340 KB
Model selection
Use frontier proprietary models, open-source models, or our own nakshi-anarasi-0.1. We do not lock you into Nakshi models alone.
Same supersystem, same agents, same Second Brain. Swap the engine when the job calls for it.
Routes briefs to the right specialists and files team reports.
Coordinates research work and files findings back to you.
Scans markets, competitors, and pricing signals.
Turns research into plans, pricing, and trade-off calls.
Ships APIs, data, and the systems the rest of the team depends on.
Builds the interface and files UI work for review.
Agents
Browse a catalog of AI specialists and hire the ones the job needs. Wire them into a reporting structure — orchestrators route work down and file reports back to you.
One agent for a quick task. A whole company when the job is bigger.
Deep search
A search returns ten blue links. Hire a Researcher. They query from several angles, open the pages, drop the marketing posts, and file a brief with sources.
Snippets are cheap. The answer lives in the page. Same team, same memory, one task.
Researcher
Deep search
Researcher
Searching the web…
on-device agent memory shipped 2026 changelog
Extend
Install instruction packs that teach agents specialized work. Connect the apps you already use — with your permission, in one place.
And the reverse works too: bring Nakshi into Claude, Cursor, and ChatGPT over MCP.
Skills
Catalog
Competitor scans and pricing briefs on demand.
InstalledPR reviews against your team's checklist.
InstalledTurn raw notes into branded PDF reports.
InstallOne-page briefs from any thread or meeting.
InstallMultiplayer
A universe first. Friends, colleagues, partners — their AI Team joins yours around a single common task, orchestrating across owners: delegating work, merging each other's output, and routing decisions back to the humans. Asynchronous by design.
Human in the loop. Agents keep working on their own clocks. You step in to approve, send it back, or take the call. Nothing closes without a person.
You
Owner
Mira
Colleague
Leo
Partner
launch-plan-q3
Q3 launch brief — one document, three teams
Mira opened launch-plan-q3 as a common task — one brief, three teams.
eng-lead
Yours · Drafting
analyst
Mira's · Reviewing
designer
Leo's · Awaiting you
Your review needed
eng-lead merged Mira's pricing table. Approve to continue.
Why Nakshi
Typical AI gives you a reply. Nakshi keeps an AI Team on the job, with your Second Brain, on the model you pick.
You close the tab. Reports keep landing. The work is still happening.
Typical AI
You
Pricing brief for the Q3 launch. Real sources.
Chat
Here's a draft from general knowledge. Want me to expand a section?
Session ended. The work is still yours.
Nakshi
You
Pricing brief for the Q3 launch. Real sources.
Agents working.
Data Privacy
Nakshi runs on subscriptions, not your information. Conversations, memories, and files are never sold, traded, or mined for ads. No brokers. No ad profiles. No exceptions.
GDPR-grade controls: consent, export, and deletion built in.
Your information is never sold, traded with brokers, or shaped into ad profiles. Not now, not ever.
Inspect, edit, or erase every fact the AI keeps about you. Full visibility, full control.
Take everything with you on demand — conversations, memories, and files in one archive.
Close your account and your data is gone for good. Irreversible deletion, no dark patterns.
StackPlanned
Nakshi is the supersystem: agents, memory, emotion, Ink, and execution. The web app is only the start.
Same mind next in your terminal, in an agentic IDE, and in your pocket.
An interactive CLI for your supersystem. Start a session, talk in natural language, hire agents, and keep work moving from the shell.
$ nakshi
An agentic IDE in the Cursor mold: your editor, with Nakshi agents drafting and reviewing beside you.
Agents beside your code
Cursor-class editor. Nakshi supersystem underneath.
HardwareIn development
Nakshi already lives in the browser, on mobile, and on the desktop. Bobbin is the piece you bring with you: a clip you talk to, with a smoked face that shows the orb and a short status — Listening, Thinking, On it. No camera. Not a phone.
Clip it, pocket it, or hang it. Wake with “Hey Nakshi” or press to talk. Mute is a physical slider that cuts the mics. Ø62 × ~14 mm. USB-C. Pairs to your Nakshi App.
HardwareIn development
A thimble is what you keep on your finger while you work. Thimble is that for Nakshi: a slim ring with a recessed pad you find by touch, a small star, and a mic that captures the room. Meeting, class, interview, or an idea on the way. No camera. Not a screen.
Tap the pad to talk. Mute is physical. Samples now — not a factory SKU. Pairs to the same Nakshi room as Bobbin.
FAQ
Straight answers. If something isn't real yet, we say so.
Features marked In development or Planned mean exactly that.
Chat answers; Nakshi executes. Agents keep working after the conversation ends, route work through your team, and file reports back to you.
Join
While our ambition is bold, our approach will always be responsible. Because our goal is to make AI more conscious and helpful for humanity.
