Everything you need to chat, search, upload, and manage your credits.
Not sure which model to choose? Start with a Daily Fast model and move up to Frontier Pro only for harder tasks.
Click your avatar in the sidebar (bottom-left). Your current balance is shown in the menu.
+ Top Up.| Top-up | Credits |
|---|---|
| 150 KES | 500,000 |
| 300 KES | 1,000,000 |
| 550 KES | 2,000,000 |
| 1,300 KES | 5,000,000 |
| Tier | Models | Best for |
|---|---|---|
| Daily Fast | DeepSeek V4.1 Flash, Gemini Flash, GPT-6 Luna | Everyday questions, drafting, summaries, quick code help. Lowest cost per message. |
| Frontier Pro | GPT-6 Sol, Claude Sonnet 5.5 | Complex reasoning, long-form writing, difficult coding and analysis. |
| Deep Research | Full-context investigation with high reasoning effort | Thorough multi-step investigations. Uses more credits, so use it when depth matters. |
!clearEvery message re-sends the earlier conversation to the model, so long chats cost more per turn. Type !clear at the very start of a prompt, or send it on its own, to prune earlier context from upstream requests.
!clear alone: everything before it is dropped from future requests.!clear Now summarise this text...: the context is cleared and your new text is kept.Your chat history remains visible on screen. Only what is sent to the model is trimmed.
Radvane integrates Tavily web search. When enabled in a chat, the model fetches current web results and answers with citations so you can verify sources.
Cost: 8,000 credits per basic search, on top of the normal token cost of the reply.
There is just one attach button (the paperclip). Pick your file and Radvane decides automatically how to read it, with no menu and nothing to choose:
| Your file | What Radvane does | Why |
|---|---|---|
| Documents with readable text (Word, typed PDFs, PowerPoint, Excel, EPUB, long text files) | Indexed for File Search: only the passages relevant to each question are sent to the model. | Cheaper on long documents, and the whole file does not have to fit in the chat. |
| Short text files and very short PDFs | Read in full, directly in the chat. | Quicker than indexing when the file is tiny. |
| Images, photos and scanned PDFs (no selectable text) | Sent straight to the AI model, which looks at the pages like a human would. | There is no text to extract, so the model needs to see the original. |
| Type | Formats | How it is handled |
|---|---|---|
| Images | PNG, JPG, WebP, GIF, HEIC, SVG | Seen directly by vision-capable models. |
| Documents | PDF, DOCX, XLSX, PPTX, EPUB, RTF | Text is extracted and read. Scanned or image-only PDFs have no text to extract, so Radvane automatically sends them to the AI as page images. Old .doc and .xls files: save as DOCX/XLSX first. |
| Text & code | TXT, MD, CSV, JSON, XML, YAML, HTML, SQL, Python, JavaScript, TypeScript, shell, and most other source files | Read as plain text. |
| Audio | MP3, WAV, M4A, WebM, OGG, FLAC | Transcribed to text first, then read by the model. Transcription is billed at 100 credits per second of audio (a 1-minute clip ≈ 6,000 credits). |
| ZIP archives | ZIP | Used for chat history import (see Import) rather than for reading. |
Radvane can remember useful details across chats (your preferences, projects, style). Manage it under Settings → Personalization, where you can view, edit, or delete memories and turn the feature off.
Open Radvane in one tap, in its own full-screen window with no browser bar. Open radvane in your browser and sign in first, then follow the steps for your device.
Using Chrome or another browser on iPhone? Tap Share → Add to Home Screen, or open Radvane in Safari.
Under every finished Radvane reply you'll see a Download bar. One click saves the reply as a PDF, a Word (.docx) file or Markdown; replies containing a table can also be saved as an Excel (.xlsx) sheet. Exports are free, don't use credits, and nothing is stored on our servers.
Bring your history from another AI platform in a few minutes. Imported chats keep every message, its order, branches, timestamps and titles, and you can continue them with any Radvane model.
| Coming from | How to export | What to upload to Radvane |
|---|---|---|
| ChatGPT (OpenAI) | ChatGPT → profile → Settings → Data controls → Export data. OpenAI emails you a ZIP link. | The ZIP file as-is (or the conversations.json inside). Large accounts that have several conversations-000.json files are merged automatically. |
| Claude (Anthropic) | Claude → Settings → Privacy → Export data. Anthropic emails you a ZIP link. | The ZIP file as-is (or the conversations.json inside). |
| Chatbot UI | Export chats from its settings. | The exported .json file. |
| LibreChat (or another Radvane account) | Settings → Data → Export a conversation. | The exported .json file. |
Settings → Data (Data & Privacy) → Import..zip or .json file. A confirmation appears and your chats show up in the sidebar (refresh the page if they do not appear straight away).!clear or start a fresh chat and paste only the part you need.Settings → Personalization (Memory) and add it as a memory entry, or just tell Radvane in a chat: "Remember that I prefer ... ". Radvane applies it to all future chats.Instructions. Upload the reference files to the agent (they are indexed for File Search). You can then reuse the agent like a custom GPT.Google Gemini, Microsoft Copilot, Perplexity and others do not offer an export Radvane can read directly. Copy the conversation (or use the platform's "share/export to document" option), then attach it to a new Radvane chat with the paperclip (it is indexed automatically) or paste the key parts and ask Radvane to continue from there.
conversations.json; for ChatGPT, do not upload chat.html.conversations*.json files one at a time.[email protected] with the platform name (do not attach your export).50,000 token credits, added when you register.
M-Pesa, via + Top Up in your avatar menu. Amounts range from 150 KES to 1,300 KES.
The whole conversation is re-sent with each message. Start a new chat or use !clear to drop old context.
Use the Concise / Balanced / Thorough pill in the top bar next to the model name. Concise: short direct answers, light reasoning, hard output cap (~4,000 tokens) — cheapest. Balanced: automatic (default). Thorough: deep reasoning and complete answers — can cost much more. Your choice is saved to your account and applies to all models.
Daily Fast for most things, Frontier Pro for demanding tasks, Deep Research for in-depth investigations.
No. There is a single attach button. Documents with readable text (Word, typed PDFs, etc.) are indexed for File Search so only relevant parts are used; images and scanned PDFs are sent straight to the AI model.
8,000 credits per basic search.
Yes. Upload the export ZIP (or its conversations.json) in Settings → Data → Import. See Switching to Radvane for all platforms, custom instructions and custom GPTs.
No. Credits are only used when you send new messages in an imported chat.
Yes. Create an Agent, paste the original instructions and upload the reference files. See Switching to Radvane.
Go to Settings → Personalization to delete memories or disable the feature.
Verify important information, especially for legal, medical, or financial matters. Use web search for citations.
Contact us at [email protected]. Please include your account email and a short description of the problem.