16 features
The right model, for every request.
Oupi automatically analyzes your message and selects the most suitable AI model. A coding question? DeepSeek or Claude will be preferred. Creative writing? GPT-5.2 takes over. Image analysis? Gemini 2.5 Pro is activated. You no longer need to know which model to use — Oupi does it for you.
💡 How does it work?
The Auto Model Selector uses a multi-criteria scoring system: task type (writing, code, analysis, translation…), estimated complexity, language, and file presence. All in under 100ms, transparent to the user.
"AI that understands your business. Not just another tool — your next competitive advantage."
"Your teams think. Our AI executes. Together, you go 10x faster."
GPT-5.2 vs Claude Opus 4.6 — in one click.
Query multiple models simultaneously on the same request and compare their responses side by side. Ideal for validating information, choosing the best writing style, or testing the reliability of a technical response. You can compare up to 4 models in parallel.
💡 How does it work?
Activate "Compare" mode in the chat bar, select your models, and send your message. Responses are displayed in columns with response time, token usage, and cost indicators.
Migrate from ChatGPT, Claude and others in 2 clicks.
Have months of conversations on ChatGPT, Claude or Gemini? Don't lose anything. Oupi imports your complete histories with metadata (dates, models used, attachments). Your conversations are restructured, indexed and immediately searchable.
💡 How does it work?
Go to Settings → Import. Upload your JSON export (ChatGPT) or connect your account directly. The import is end-to-end encrypted and data never leaves Europe.
ChatGPT
1,247 conversations · JSON export
Claude (Anthropic)
832 conversations · Direct connection
Gemini (Google)
415 conversations · JSON export
Privacy Mode
Protected session · No data stored
0
Data stored
0
Server logs
0
AI training
100% anonymized conversations.
Enable Privacy Mode so your conversations are never stored or used for model training. No trace, no logs, no memory. Ideal for sensitive topics, medical or legal consultations, and any conversations you want to keep strictly private.
💡 How does it work?
A simple toggle in the chat bar activates Privacy Mode. The green shield icon confirms the session is protected. Data is processed in volatile memory only and deleted when the conversation is closed.
Your sensitive data never leaves your walls.
Privacy Protection goes even further than Privacy Mode. Oupi automatically detects and tokenizes sensitive data before it reaches AI models: email addresses, phone numbers, financial amounts, client names, contractual clauses, medical data. Models only see anonymous tokens — and Oupi reconstructs the real data only in your interface.
💡 How does it work?
Tokenization relies on a European NER (Named Entity Recognition) engine that identifies 15+ categories of sensitive data. Tokens replace data before sending to the model (e.g.: [EMAIL_001], [AMOUNT_003]). Upon reception, Oupi re-substitutes the real content client-side, without ever exposing it to third-party AI servers.
Original message:
"Send the contract to [email protected] for an amount of 45 000€"
Sent to AI model:
"Send the contract to [EMAIL_001] for an amount of [MONTANT_001]"
Available models
GPT-5.2, Opus 4.6, Gemini, DeepSeek, Mistral… all accessible.
No more juggling between ChatGPT, Claude and Gemini. Oupi brings all leading models into a single interface. Switch models mid-conversation without losing context. Updates are near-instant: when a new model launches, it's available in Oupi within 24 to 48 hours.
💡 How does it work?
Oupi interfaces directly with official APIs from OpenAI, Anthropic, Google, Meta, Mistral, DeepSeek, xAI, Cohere and others. Each model is optimized with specific system prompts and parameters to ensure the best performance.
Edit AI responses, directly in the chat.
Canvas Mode transforms an AI response into an editable document. Modify text, reorganize sections, adjust the tone — all without leaving the chat. Once satisfied, export in your preferred format: Word (.docx), PDF, Markdown, HTML or plain text. Perfect for reports, articles, emails and professional documents.
💡 How does it work?
Click the Canvas icon on any response. The editor opens in a side panel with syntax highlighting, rich formatting, and the ability to ask AI to rework selected passages.
👤 Profile
CTO at TechVision SAS. Expert in Laravel, Vue.js, AWS cloud architecture.
✍️ Style
Prefers concise answers with code examples. Informal tone. No emojis.
🏢 Context
B2B SaaS startup, 25 employees. Stack: Laravel 11, Vue 3, PostgreSQL, Redis.
Increasingly relevant answers, conversation after conversation.
Teach Oupi your personal and professional context: your role, industry, writing preferences, tech stack, goals. The more Oupi knows you, the more precise and tailored its responses become. Knowledge persists between conversations and can be updated at any time.
💡 How does it work?
Go to Settings → Knowledge. Add information as free text or structured data. Oupi automatically injects relevant context into each conversation without consuming additional visible tokens.
Pre-save your prompts to save time.
Create, organize and reuse your most effective prompts. Categorize by use (writing, code, analysis, translation…), add dynamic variables ({{client_name}}, {{topic}}…), and launch them in one click. Share your best prompts with your team to standardize practices.
💡 How does it work?
From the chat bar, type "/" to access your library. Each prompt can contain variables that Oupi will ask you to fill in before sending. Prompts can also be triggered by keyboard shortcuts.
Write an email to {{recipient}} about {{topic}} in a {{tone}} tone…
Analyze main clauses, risks and recommendations…
Review this code: bugs, performance, security, best practices…
Organize your chats by project, client or topic.
No more endless scrolling through conversation lists. Create folders and subfolders to organize your chats by project, client, department or topic. Drag and drop conversations, use colored tags, and find any exchange with full-text search.
💡 How does it work?
The left sidebar shows your folder tree. Right-click to create, rename, move. Folders support auto-sorting: Oupi can automatically suggest which folder to file a new chat in.
Specialized AI experts, ready to use.
Access pre-configured chatbots optimized for specific domains: legal, finance, medical, technical, marketing, HR. Each chatbot knows industry vocabulary, expected formats and domain best practices. The legal chatbot writes in conclusion format, the finance chatbot generates structured tables, the medical chatbot respects clinical terminology.
💡 How does it work?
Select a specialized chatbot from the chat home page or via the selector at the top of the conversation. Each chatbot uses an expert system prompt, an optimized model and domain-specific tools.
Specialized Chatbots
Legal
Contracts, GDPR, corporate law
Finance
Analysis, reporting, forecasting
Medical
Clinical terminology, protocols
Technical
Code, debugging, architecture
Marketing
SEO, copywriting, strategy
RH
Recruitment, job descriptions
✦ Apple announced several new developments in February 2026:
1. Apple Intelligence 2.0 — New version of the AI engine…
2. MacBook Pro M5 — M5 chip with Neural Engine…
Up-to-date answers, connected to the web.
AI is no longer limited to its training data. With Live Data, Oupi performs real-time web searches via Google and Perplexity to enrich its responses with the most recent information. News, stock prices, weather, sports results, new laws — everything is accessible.
💡 How does it work?
Enable Live Data in the chat options or directly ask a current events question. Oupi automatically detects when a web search is needed, launches the query, synthesizes results and cites its sources.
Connect your internal data for tailored answers.
RAG (Retrieval-Augmented Generation) connects your knowledge bases directly to the chat. Import your internal documents, CRM, technical databases — and ask questions in natural language. "Who is the last signed client?", "What is my latest Amazon purchase?", "What does our HR policy say about leave?" — Oupi searches your data and answers precisely, citing its sources.
💡 How does it work?
Go to Settings → RAG → Connect a source. Upload files (PDF, CSV, JSON, SQL) or connect services (Google Drive, Notion, Confluence). Data is vectorized and indexed locally on our European servers. Searches use a hybrid algorithm (semantic + keywords) for maximum relevance.
✦ According to your CRM, the last signed client is TechVision SAS on February 3, 2026, for a contract of €48,000/year (Enterprise license).
✦ Your next meeting is the "Sprint Review" at 3:30 PM (in 1h22).
✓ I created the appointment "Meeting with Marie" tomorrow at 2:00 PM in your Google Calendar. View ↗
Your calendar, your tools, your services — from the chat.
MCP (Model Context Protocol) allows Oupi to interact directly with your professional tools. Check your Google Calendar, create appointments, send emails, create Jira tasks, query Slack — all in natural language from the chat. "What time is my next meeting?", "Create an appointment tomorrow at 2pm with Marie" — Oupi executes.
💡 How does it work?
Connect your services in Settings → Integrations. Oupi supports 120+ connectors: Google Workspace, Microsoft 365, Slack, Jira, Notion, Salesforce, HubSpot, GitHub and many more. MCPs work in read AND write mode, with granular permissions you control.
Speak directly to AI, it responds.
No more typing. Enable voice mode and converse naturally with Oupi. Voice recognition works in real time with near-zero latency. Oupi understands French, English, Spanish, German and 30+ other languages. Text-to-speech delivers responses out loud with a natural voice.
💡 How does it work?
Click the microphone icon in the chat bar or use the shortcut (Ctrl+M). Transcription displays in real time. Oupi uses Whisper for recognition and ElevenLabs for synthesis.
Voice Mode active
Speak naturally, Oupi transcribes and responds
Create visuals and advanced interactions, in the chat.
Oupi is not limited to text. Generate images with DALL-E 3, Midjourney or Stable Diffusion directly in the chat. Create charts, diagrams, presentations, interactive tables. One multi-AI environment for all your creations, without switching tools.
💡 How does it work?
Describe the desired image in natural language, or use the /image prefix to force visual generation. Images can be retouched, varied and exported in high resolution. Charts are generated as interactive SVGs.