The Origin Story
Rilobase was born out of pure frustration with modern AI models.
You ask an AI a hard question, and it gives you a shallow, generic answer. If you ask it to stress-test that answer, it just regurgitates the same median internet opinion. If you ask it for prices or live data, it hallucinates or reads SEO-optimized blog posts written by other AIs.
We realized the problem isn't that the AI models aren't smart enough. The problem is they are guessing from outdated training data instead of looking at reality. Rilobase was built to fix this. We don't train models. We force the AI to read verified, human-written primary sources before it is allowed to speak.
Quickstart
Launch App
Click 'Launch App' and create a secure account.
Open Settings
Navigate to Settings via the sidebar profile menu.
Connect Provider
Select your AI provider (OpenAI, Anthropic, Groq, or Google) and securely connect your API key.
Start Researching
Return to the dashboard and ask a question. The verification happens automatically.
Zero-Knowledge BYOK
Unlike AI wrappers that mark up your token usage, Rilobase operates on a strict Bring Your Own Key (BYOK) model. You pay the AI provider directly for the compute. We just provide the high-speed routing and research pipeline.
Zero-Knowledge Architecture
Your API keys are encrypted at rest using AES-256. They are decrypted in memory only for the milliseconds it takes to forward your request, then immediately discarded. We cannot view, extract, or use your keys.
The Anti-Slop Engine
The internet is filled with "AI slop"—blogs written by AI, summarizing other AI content. If a standard AI scraper reads this, it creates diluted, generic garbage. Rilobase actively rejects this using mathematical quality filters.
Domain Blacklisting
We hard-block domains known for user-generated content and SEO farms. Wikipedia, Medium, and generic SaaS blogs are instantly discarded before scraping.
Density & Burstiness Math
Our backend analyzes the mathematical structure of scraped text. It counts data points (numbers, dates) and measures sentence length variation. If the text lacks data and has uniform structure, it's flagged as AI slop and rejected.
The Global Fact Flywheel
Standard caches match exact strings. Rilobase uses an intelligent memory engine that understands the *meaning* of your question. If User A asks "Price of RTX 4090" and User B asks "How much is a 4090", Rilobase recognizes the intent is the same and returns the verified answer instantly.
This creates a global flywheel: the more the platform is used, the faster and cheaper it gets for everyone. Generative prompts (like "write an essay") bypass this cache to ensure you always get a unique, creative output, but they still use the cached verified facts as their foundation.
Advanced Retrieval Engine
Many modern websites rely on JavaScript to render their data, meaning standard API scrapers see nothing but blank code. If your AI can't read the primary source, it hallucinates.
Rilobase utilizes an advanced rendering engine that executes web pages dynamically, allowing us to read live, structured data that standard AI wrappers cannot see. We strictly adhere to ethical scraping guidelines and `robots.txt` directives, ensuring we only pull data from sources that permit it, while bypassing the SEO spam that clutters standard search results.
Data & Privacy
Your chat history is tied to your account and is not shared. You can wipe your chat history at any time from the Data & Privacy tab in Settings.
Sealed Frontend: All database interactions are routed securely through our application server. Your authentication tokens and backend URLs are never exposed directly to the browser, ensuring a secure, sealed environment.