While working on a college project heavily involving SLMs (Small Language Models), I realized something: smaller, low-cost models are incredibly powerful if you give them the exact right context and architecture.

That realization is why I spent this summer building ManjuAi: https://manjuai.com

I wanted this to be completely friction-free for other devs, so I set up a zero-auth sandbox—no signup or email required. You get 5 free queries a day directly on the landing page to break it and test it. Free users get to choose from about 39 free/low-cost models (including DeepSeek V3/R1, GPT mini/nano models, Gemini Pro previews, and Claude Haiku).

### 🛠️ What ManjuAi Solves:

* 1. CodeOn (ZIP Context Explorer): Chatbots fail when they assume your architecture. As a developer, I used to upload snippets of code where the bug was, but without the full project context, the model just guessed and gave wrong answers. ManjuAi lets you upload a full project `.zip` file, unpacks it into an in-browser file explorer, and lets you query and edit files directly so you don't have to jump back and forth to your IDE. You can even prompt it to "create a simple HTML file" and watch it populate a new file into the tree automatically.

* 2. Controlled Deep Research: Most research agents dump a massive report of whatever length they feel like. ManjuAi lets you select the exact page count you want. A multilayered scraping pipeline hunts down the best resources and structures the output exactly to your constraints—ensuring deep research done right and at low cost.

* 3. Chat & Source Synthesis: Add up to 5 attachments (links, photos, PDFs, or any other document) to synthesize deep reports or generate audio-ready scripts. You can access top web searches, links, videos, and news directly in the workspace topbar, or instantly convert your ongoing conversation into a clean markdown report.

### 💰 Pricing

If you want to scale up your usage, we have a Pro tier for just $5.30 (500 INR) which increases your limits by 5x and grants access to premium flagship models (like the latest GPT, Gemini, and Claude Sonnet suites).

The website is live right now. Please drag a project folder ZIP into it, run a research loop, and let me know what breaks or what features you want to see next!