About
An independent AI workspace, built by one person
PhantomAI isn't a venture-backed company with a "team" — it's a solo-built product. I think that's worth saying plainly, because I'd rather you trust exactly what this is than trust a polished story about something bigger.
Simanta Pratim Das
Founder & Developer
Simanta is an independent AI engineer based in Guwahati, India, building PhantomAI as a solo project — designing the product, the interface, and the AI pipeline end to end.
Why I built this
The first wave of AI tools were astonishing in capability but clumsy in practice — scattered across a dozen tabs, forgetful between sessions, and often vague about where an answer came from or how your data was handled. I wanted one place that felt as considered as the best consumer software, but with the depth real work demands. So I built it, end to end: the chat experience, the voice mode, the live canvas, and the backend pipeline that ties it together.
This is not a claim that PhantomAI rivals the resources of a large AI lab — it doesn't, and I won't pretend otherwise. What I can offer is a product built and maintained by someone who uses it daily, responds to feedback directly, and has no incentive to hide how it actually works.
How PhantomAI actually works
No black box — here's the real technical picture, so you know what you're using.
The chat & AI pipeline
PhantomAI runs on a Next.js frontend with an Express backend, streaming responses over server-sent events. Requests go through a small agent pipeline — a planner classifies the task, a reasoning step calls the model, and a formatter cleans the output before it reaches your screen.
The models
Chat responses are generated by open-weight large language models served through Groq's inference infrastructure. Model choice depends on the task: faster models handle quick questions, larger ones handle deeper reasoning and code.
Storage & auth
Accounts and sign-in are handled by Clerk. Conversation history is stored in MongoDB and tied to your account, so you can pick up a chat on a different device — and delete it permanently whenever you choose.
What I'm optimizing for
A single, well-made AI workspace that's genuinely useful for everyday work — fast, private, and honest about what it is: a product built by one developer, not a faceless company.
Where this is headed
I ship improvements continuously based on how I and early users actually use the product — more model choice, better memory, and a canvas that keeps closing the gap with a real IDE.
What I value
Usefulness first
Every feature has to earn its place by saving you real time or helping you think more clearly. I build what I would actually reach for myself, not demo material.
Privacy by default
Your conversations belong to you. I minimize what gets collected, I'm transparent about how it's used, and you can export or delete your data whenever you want.
Honest about limits
AI is powerful but imperfect, and so is a one-person product. I'd rather tell you what PhantomAI can't do yet than oversell what it can.
Craft and clarity
From the chat interface to the writing on this page, I sweat the details myself. Good tools should feel calm, fast, and obvious to use.
Have a question, or found a bug?
Reach out — messages go directly to me, not a support queue.