Konnect.ai User Manual
A comprehensive guide to using Konnect.ai - your multi-model AI platform.
1. Getting Started
Creating an Account
- Visit konnect.ai and click "Get Started"
- Enter your email and create a password
- Verify your email address
- You'll be redirected to your dashboard
Subscription Tiers
| Tier | Modes | Queries | Models |
|---|---|---|---|
| Free | Smart Chat | 50/month | GPT-4o, Claude Haiku, Gemini Pro |
| Pro | + Ensemble, Debate, Perspectives | 500/month | + GPT-5.2, Claude Sonnet 4.5 |
| Power | + Council, Red/Blue Team | Unlimited | All models |
2. Dashboard Overview
The dashboard is your home base after logging in. Here you can:
- Start a new chat - Click the "New Chat" button
- Choose a mode - Select from Smart Chat, Ensemble, Debate, Perspectives, or Red/Blue Team
- View recent sessions - Access your conversation history
- See usage stats - Track your query usage for the month
Tip: Click on any mode card to start a new chat with that mode pre-selected.
4. Smart Chat Mode
Smart Chat automatically routes your query to the most suitable AI model based on the type of question you're asking.
How It Works
- Type your question in the chat input
- Press Enter or click Send
- The system analyzes your query and selects the best model
- You receive a response from that model
Best For
- Quick questions with clear answers
- General knowledge queries
- When you don't need multiple perspectives
- Maximizing speed and minimizing cost
5. Ensemble Mode
Ensemble mode queries multiple AI models simultaneously and synthesizes their responses into a single, comprehensive answer.
Getting Started
- Select "Ensemble" mode from the mode dropdown in the chat input
- By default, GPT-5.2 and Claude Sonnet 4.5 are pre-selected
- Aggregation method defaults to Synthesis
- Type your question and press Enter
Selecting Models
Click the Add button to open the model picker:
- Select up to 5 models
- Mix providers (OpenAI, Anthropic, Google) for diverse perspectives
- See estimated cost per query based on selected models
- Models show their tier requirement (Free, Pro, Power)
Aggregation Methods
Synthesis (Default)
An AI reads all responses and creates a unified answer combining the best insights. Use for most questions.
Best of N
Selects the single best response based on quality. Use when you want one definitive answer.
Consensus
Highlights where models agree and flags disagreements. Use for fact-checking.
Union
Combines all unique perspectives. Use for comprehensive coverage of a topic.
Reading the Response
After the response loads, you'll see:
- Synthesized tab: The combined/aggregated response (shown by default)
- Model tabs: Click individual model names to see their full responses
- Consensus section: Shows where models agreed or disagreed
- Latency info: How long each model took to respond
Follow-up Questions
When you ask a follow-up question in an ensemble conversation:
- The same models are queried again
- Each model receives the synthesized response as context (not individual responses)
- Models don't see what other models said individually
- This keeps the conversation coherent and context manageable
6. Debate Mode
Debate mode pits two AI models against each other in a structured argument, with a third model acting as judge.
Debate Structure
Pro (Argues FOR)
Default: GPT-5.2
Con (Argues AGAINST)
Default: Claude Sonnet 4.5
Judge (Evaluates)
Default: GPT-4o
Debate Rounds
- Round 1 - Opening Statements: Each side presents their initial position
- Round 2 - Rebuttals: Each side responds to the other's arguments
- Round 3 - Closing Arguments: Final appeals and summaries
- Verdict: Judge evaluates arguments and declares a winner
Starting a Debate
- Select "Debate" mode from the dropdown
- Optionally change the Pro, Con, or Judge models
- Enter a topic that has two clear sides (e.g., "Is remote work better than office work?")
- Press Enter and watch the debate unfold
After the Debate
- Rematch: Run the same debate again (results may vary)
- Follow-up: Ask questions about specific arguments
- Share: Make the debate public with a shareable link
- New Topic: Start a fresh debate with a different topic
Tips for Good Debate Topics
- Frame as a clear statement or question with two obvious sides
- Avoid loaded or biased phrasing
- Good: "Should companies require employees to return to office?"
- Bad: "Why is remote work obviously better?" (biased)
7. Perspectives Mode
Perspectives mode provides three distinct viewpoints on any topic—Optimist, Realist, and Pessimist—then synthesizes them into a balanced conclusion.
The Three Perspectives
Optimist
Focuses on opportunities, benefits, and positive outcomes. Default: GPT-5.2
Realist
Provides objective, evidence-based analysis without bias. Default: Claude Sonnet 4.5
Pessimist
Identifies risks, challenges, and potential downsides. Default: Gemini 2.5
Using Perspectives Mode
- Select "Perspectives" mode from the dropdown
- Optionally change the models for each perspective (click on the role labels)
- Enter a topic or question that benefits from multiple viewpoints
- Watch as each perspective completes in parallel, followed by the synthesis
Reading the Response
- Three perspective cards: Each shows the viewpoint with key points
- Synthesis panel: Balanced conclusion with areas of agreement and uncertainties
- Progress indicator: Shows how many perspectives have completed (e.g., "2/3 perspectives")
Best For
- Major life or business decisions
- Investment and financial analysis
- Risk assessment for projects
- Career or educational choices
- Technology adoption decisions
Follow-up Questions
After a perspectives analysis completes, click "Ask About This" to enter follow-up mode:
- Ask questions about specific perspectives or the synthesis
- Get clarification on areas of agreement or uncertainties
- Explore the recommendation in more depth
- The follow-up uses context from all three perspectives for balanced answers
Tip: Ask follow-up questions like "Which perspective best addresses long-term risks?" or "What do all three agree on regarding timing?" to get targeted insights.
8. Red/Blue Team Mode
Red/Blue Team mode performs adversarial security analysis. A "Red Team" model attacks your content to find vulnerabilities, while a "Blue Team" model defends and hardens it.
The Two Teams
Red Team (Attacker)
Analyzes content to find security vulnerabilities, weaknesses, and attack vectors.
Blue Team (Defender)
Evaluates vulnerabilities, disputes false positives, and provides hardened content.
Content Types
- Code: Find security vulnerabilities in source code
- Documents: Analyze contracts, policies, or legal text
- Prompts: Test AI prompts for injection vulnerabilities
Intensity Levels
- Light: Quick scan for obvious issues
- Standard: Balanced analysis (default)
- Thorough: Deep, comprehensive security review
Reading the Response
- Vulnerability list: Issues found, categorized by severity (Critical, High, Medium, Low)
- Statistics: Summary of found, resolved, and disputed vulnerabilities
- Confidence score: How confident the analysis is (0-100%)
- Hardened content: Your content with vulnerabilities fixed
9. Personas
Personas are custom configurations that modify how AI models respond. They add a system prompt that shapes the model's behavior, tone, and expertise.
Creating a Persona
- Go to Personas in the sidebar
- Click New Persona
- Enter a name (e.g., "Code Reviewer")
- Write a description of what it does
- Write the system prompt that defines its behavior
- Click Save
Using Personas
In Ensemble or Debate mode:
- Click "Add" to add a model slot
- Select a model from the dropdown
- Select a persona to apply (optional)
- The persona's system prompt will guide that model's responses
Example Personas
Code Reviewer
"You are a senior software engineer conducting code reviews. Focus on code quality, potential bugs, security vulnerabilities, and best practices."
Devil's Advocate
"You challenge assumptions and play devil's advocate. Question ideas constructively and present alternative viewpoints."
10. Settings & Preferences
Access settings from the sidebar to configure your Konnect.ai experience.
General
- Profile name and avatar
- Email preferences
- Default model preferences for each mode
Billing
- View current subscription tier
- Track monthly usage
- Upgrade or downgrade plans
- Manage payment methods
- Download invoices
Security
- Change password
- Enable two-factor authentication
- Manage API keys
- View active sessions
11. Edge Cases & Behaviors
Ensemble Mode
What if one model fails?
If a model times out or returns an error, the synthesis continues with the remaining models. You'll see an error indicator on that model's tab. If ALL models fail, you'll see an error message.
What if models completely disagree?
The synthesized response will acknowledge the disagreement and present both viewpoints. The Consensus section will highlight the specific points of disagreement.
Can I change models mid-conversation?
No. Once you send your first message, the model selection is locked for that conversation. This ensures consistency. Start a new chat to use different models.
What if I select only one model?
You need at least 2 models for ensemble. With one model, it behaves like Smart Chat.
Why is ensemble slower than Smart Chat?
Ensemble queries multiple models in parallel, waits for all responses, then runs a synthesis step. With web search enabled, each model may also make multiple API calls. Expect 10-30 seconds for complex queries.
Debate Mode
What if a debate times out mid-round?
You'll see an "Incomplete" verdict. The debate shows as much as was completed. You can still use Rematch to try again.
Can I use the same model for Pro and Con?
Yes, but using different models produces more interesting debates with genuine differences in reasoning style.
Does the judge model affect the outcome?
Yes. Different judge models may weigh arguments differently. Try different judges with Rematch to see how verdicts change.
What topics work poorly for debates?
Factual questions with clear right answers ("What is 2+2?") don't make good debates. Also avoid topics that are too personal or lack a clear opposing position.
General
What happens if I exceed my monthly quota?
You'll see a message asking you to upgrade or wait for the next billing cycle. Queries won't be processed until you have available quota.
Are my conversations stored?
Yes, your conversations are stored and accessible from the sidebar. You can delete individual sessions. Public sessions remain accessible via their share link.
Can I use Konnect.ai offline?
No. Konnect.ai requires an internet connection to query AI models.
12. Troubleshooting
"Response generation failed"
Causes: Model API timeout, rate limiting, or service outage.
Solution: Wait a moment and try again. If persistent, try different models or check our status page.
Ensemble shows only synthesized response, no model tabs
Causes: Individual model responses may have failed or returned empty.
Solution: Check if you have sufficient quota. Try the query again or select different models.
Debate stuck on "Loading..."
Causes: Debates run 3 rounds with 3 models - this takes time.
Solution: Wait up to 60 seconds. If still stuck, refresh and check your session history - the debate may have completed.
Can't access Pro features
Causes: Free tier account or subscription expired.
Solution: Go to Settings → Billing to upgrade your subscription.
Session not saving
Causes: Browser storage issues or network disconnection.
Solution: Check your internet connection. Try clearing browser cache and signing in again.
13. Frequently Asked Questions
Q: Is my data used to train AI models?
No. Your conversations are not used to train any AI models. Data is stored only for your session history.
Q: Which model is best?
It depends on your use case. GPT-5.2 excels at reasoning, Claude Sonnet 4.5 at nuanced analysis, and Gemini at multimodal tasks. That's why ensemble mode is powerful - it combines their strengths.
Q: Can I use Konnect.ai via API?
Yes! Generate an API key in Settings → Security. See the API Reference documentation for endpoints and examples.
Q: How is pricing calculated for ensemble?
One ensemble query counts as one query toward your quota, regardless of how many models are used. However, the computational cost to us is higher, which is why ensemble requires a Pro subscription.
Q: Can I export my conversation history?
Currently, you can share public sessions via link. Full export functionality is coming soon.
Q: Is there a mobile app?
Not yet, but the web interface is fully responsive and works well on mobile browsers.
Need More Help?
Can't find what you're looking for? We're here to help.