Most AI tools follow a straight line: you ask a question, you get an answer, you move on. The research ends where the response ends.
Citra AI doesn't work that way. We designed it as a recursive loop — a system where every answer is a doorway into deeper understanding. You don't just get information from Citra. You build memory as you work, and the more you explore, the smarter your workspace becomes.
Here's why we built it this way, and what it means for how you actually do research.
The Problem With Linear Research
Think about how research usually works with AI tools. You upload a document to ChatGPT. You ask a question. You get a summary. Done. If you want to go deeper into a specific section, you re-prompt. If you want to explore a related concept, you start a new conversation. If you want to see how ideas connect across multiple documents, you're on your own.
The workflow is flat. There's no depth. No ability to drill into a citation, explore a source, pivot to a related question, visualize relationships, and then come back to where you left off — all within the same workspace, with the same AI that already knows your full context.
That's the gap Citra AI was designed to close.
The Recursive Loop: How It Works
In Citra, research isn't a single action — it's a continuous cycle of asking, exploring, understanding, and building. Each layer feeds the next. Here's the loop in action:
Build your Vault memory
You start by uploading your project files to Citra Vault — PDFs, documents, spreadsheets, web links. You can also record meetings directly into the Vault, or let Citra browse the web and ingest research for you. This is your living knowledge base. It grows as you work.
Chat with your Vault
Ask questions about your project. The AI draws from everything in your Vault and responds with citations and references — linking back to specific pages in your PDFs, paragraphs in your documents, or internet sources Citra used (if you've enabled web research). Every answer is grounded in your actual sources.
Deep-dive into a source
See a citation that catches your eye? Click it. The source opens in Citra's built-in Reader — with the same AI sitting right beside it in a side panel. Now you're inside that specific document, and you can ask the AI: "Explain this section," "Summarize this PDF," or "Generate a diagram from this." You've gone one level deeper — from project-level research to document-level understanding.
Follow the thread wherever it leads
While exploring that document, a new question comes up. Maybe it's related to the document, maybe it's something entirely different. Ask it. Citra pulls from your Vault, from the internet, from whatever it needs to get you the answer. You're not locked into one document or one train of thought — the AI has your full project context, always.
Come back to the chat
When you're ready, return to your project chat. The context is right where you left it. Ask a few more questions. Explore another citation. Open another document in the Reader. Every time you go deep and come back, your understanding compounds.
Switch to visual mode
Sometimes linear text isn't enough. Open your documents as a knowledge graph to see entity relationships as a network. Switch to a mindmap to see hierarchies. Generate a diagram to understand a process. Each visualization type reveals something different — and each has its own side-by-side AI chat to guide your exploration.
Loop back — deeper than before
A node in your knowledge graph sparks a question. Ask it. A branch in your mindmap needs exploration. Click into it. A diagram reveals a gap in your understanding. Chat about it. Every output is an input to the next cycle. That's the recursive loop.
↓
Visualize → Knowledge Graph / Mindmap / Diagram → New Questions → Back to Chat
↓
↻ Repeat — Each pass builds deeper memory
Why Recursive? Because Understanding Isn't Linear
Real research doesn't happen in a straight line. You read a paper, it references another concept, you explore that, it connects to something in a completely different document, and suddenly the picture becomes clear in a way that no single source could have given you.
We built Citra's research this way because that's how researchers actually think:
- You need depth, not just answers. A chatbot gives you a response. Citra lets you drill into the source, question the source, and then explore what the source connects to.
- You need context that persists. When you go deep into one document and come back, the AI should still know your full project. In Citra, it does — always.
- You need multiple lenses. Text is one way to understand. Knowledge graphs show you the network of entities and relationships. Mindmaps show hierarchy. Diagrams show processes. Sometimes you need to see the structure, not just read about it.
- You need one workspace. The moment you switch tools — from a chat app to a PDF reader to a diagramming tool — you lose context. Everything in Citra happens in one place, with one AI that knows everything you've fed it.
Memory That Grows as You Work
The key insight behind the recursive loop is that Citra Vault is not just storage — it's memory. Every file you upload, every meeting you record, every web page Citra ingests, every conversation you have — it all becomes part of the project's living knowledge base.
This means:
- The first question you ask draws from what you uploaded.
- The tenth question draws from your uploads, plus everything the AI has learned about your project through conversation.
- The hundredth question? The AI understands your project deeply — the terminology, the relationships, the priorities, the gaps. It gives better answers because it has deeper memory.
This is what we mean by Deep Project Memory. It's not just retrieval. It's accumulated understanding that compounds over time.
The Researcher's Advantage
For researchers, analysts, consultants, and anyone who works with complex information: Citra's recursive loop means you don't just find answers — you build understanding. Every session makes the next one more productive. Every deep-dive enriches your Vault. And when it's time to create a living presentation or executive report from your research, the AI already knows everything. One click, and your research becomes a deliverable.
What Citra Vault Holds Today
Right now, your Vault memory can be built from:
- File uploads — PDFs, documents, spreadsheets, images. The AI indexes and understands the full content.
- Meeting recordings — Record meetings directly in Citra. Transcriptions are indexed and become part of your Vault memory, searchable and citable just like any document.
- Web browsing & ingestion — Let Citra browse the internet, summarize web pages, and ingest content into your Vault. The AI brings the web to your project context.
- Conversation history — Your chats with the Vault AI build up project context over time through Deep Project Memory.
🚀 Coming Soon: Citra Desktop
Citra Desktop is coming soon — where users who have already set up permissions and privacy controls on their own infrastructure can connect their data sources directly to the research loop. Connect to SQL Server, Databricks, data lakes, and other enterprise sources, all processed locally on your machine. Your data never leaves your environment. Same recursive research loop, but now powered by your entire data infrastructure.
From Research to Output — Without Breaking the Loop
Here's the part that ties it all together: the recursive research loop doesn't end at understanding. When you're ready, the same Vault that powered your research powers your output:
- Generate presentations grounded in your Vault — every slide citable, every claim traceable.
- Generate reports that pull from the full depth of your research.
- Generate knowledge graphs, mindmaps, and diagrams that visualize what you've learned.
- Share a living link that stays current — when your Vault evolves, your outputs evolve with it.
Research, understand, create, share, update. It's all one loop. That's why we built Citra this way.