The Chat feature provides a powerful conversational AI interface powered by Google Gemini. Engage in natural conversations, get help with tasks, and leverage MCP tools for enhanced capabilities.
Overview
Chat offers:
- Natural Conversations - Fluid, context-aware dialogue
- Multiple Models - Choose between Gemini models
- MCP Integration - Enable tools for extended capabilities
- Conversation History - Manage multiple conversations
- Markdown Support - Rich text formatting in responses
Getting Started
Starting a Conversation
- Navigate to Chat in the sidebar
- Click New Chat to start a conversation
- Type your message in the input field
- Press Enter or click Send
Interface Layout
┌─────────────────────────────────────────────────────────┐
│ Sidebar │ Chat Area │
│ ───────────── │ ────────────────────────────── │
│ [+ New Chat] │ │
│ │ ┌─────────────────────────────┐ │
│ Conversations: │ │ User: Hello! │ │
│ • Chat 1 │ └─────────────────────────────┘ │
│ • Chat 2 │ │
│ • Chat 3 │ ┌─────────────────────────────┐ │
│ │ │ AI: Hi! How can I help? │ │
│ ───────────── │ └─────────────────────────────┘ │
│ Model: [Gemini] │ │
│ │ ───────────────────────────────── │
│ ───────────── │ [ Type your message... ] [Send] │
│ MCP Tools ▼ │ │
└─────────────────────────────────────────────────────────┘
Model Selection
Available Models
| Model | Description | Best For |
|---|---|---|
| Gemini 2.0 Flash | Fast, capable model | General conversations, quick tasks |
| Gemini 1.5 Pro | Advanced reasoning | Complex analysis, detailed explanations |
Changing Models
- Locate the Model selector in the sidebar
- Click to open the dropdown
- Select your preferred model
- New messages will use the selected model
Note: Model changes apply to new messages. Previous messages retain their original model context.
Conversations
Managing Conversations
- Create: Click New Chat to start fresh
- Switch: Click any conversation in the sidebar
- Delete: Hover over a conversation and click the trash icon
- Rename: Conversations auto-title based on first message
Conversation Context
Each conversation maintains context:
- Previous messages are remembered
- AI responses consider the full conversation
- Context persists across sessions (stored locally)
Example Conversation
You: Can you explain what machine learning is?
AI: Machine learning is a subset of artificial intelligence
that enables systems to learn and improve from experience
without being explicitly programmed...
You: How is it different from deep learning?
AI: Great follow-up question! Deep learning is actually a
specialized subset of machine learning that uses neural
networks with multiple layers...
MCP Tools Integration
What are MCP Tools?
MCP (Model Context Protocol) tools extend the AI's capabilities by connecting to external services:
- Databases - Query and analyze data
- File Systems - Read and write files
- APIs - Interact with external services
- Web Search - Find information online
Enabling MCP Tools
- Expand the MCP Tools panel in the sidebar
- View available connected servers
- Click to toggle servers on/off
- Enabled servers show with a checkmark
Tool Usage in Chat
When tools are enabled, the AI can:
- Recognize when a tool would help
- Call the appropriate tool
- Process the results
- Provide informed responses
Example with Web Search:
You: What are the latest news about AI regulations?
AI: [Using web search tool...]
Based on recent news, here are the key developments
in AI regulations:
1. The EU AI Act has entered its implementation phase...
2. The US has proposed new guidelines for...
3. China has updated its regulations regarding...
Sources: [links to articles]
Available Tool Types
| Tool | Capability | Example Use |
|---|---|---|
| PostgreSQL | Query databases | "Show me all orders from last month" |
| Filesystem | Read/write files | "Read the contents of config.json" |
| GitHub | Repository operations | "List open issues in the repo" |
| Web Search | Internet search | "Find recent articles about..." |
| Memory | Store/retrieve info | "Remember my preferences" |
Message Features
Markdown Support
AI responses support full markdown:
- Headers - Organize information
- Lists - Bullet and numbered lists
- Code Blocks - Syntax-highlighted code
- Tables - Structured data
- Links - Clickable URLs
Example Response:
# Quick Guide
Here's how to get started:
1. Install dependencies
2. Configure settings
3. Run the application
## Code Example
```javascript
const greeting = "Hello, World!";
console.log(greeting);
### Code Copying
For code blocks in responses:
1. Hover over the code block
2. Click the copy icon
3. Code is copied to clipboard
### Message Actions
| Action | How To | Result |
|--------|--------|--------|
| **Copy** | Click copy icon on message | Copies text to clipboard |
| **Retry** | Click retry icon | Regenerates AI response |
## Keyboard Shortcuts
| Shortcut | Action |
|----------|--------|
| `Enter` | Send message |
| `Shift + Enter` | New line |
| `↑` | Edit last message |
| `Ctrl/Cmd + K` | Focus search |
## Use Cases
### General Assistance
You: Help me write a professional email to decline a meeting invitation politely.
AI: Here's a professional email template...
### Code Help
You: How do I sort an array of objects by date in JavaScript?
AI: Here are a few approaches to sort an array by date...
const sorted = array.sort((a, b) =>
new Date(a.date) - new Date(b.date)
);
### Research & Analysis
You: Compare the pros and cons of React vs Vue for a large enterprise application.
AI: Great question! Let me break down the comparison...
| Aspect | React | Vue |
|---|---|---|
| Learning Curve | Moderate | Gentle |
| Ecosystem | Extensive | Growing |
| ... |
### Creative Writing
You: Write a short story opening about a robot discovering emotions for the first time.
AI: Unit-7 had processed 47,832 days of data before the anomaly occurred. It started as a minor deviation...
## Tips & Best Practices
### For Better Responses
1. **Be Specific** - Clear questions get better answers
2. **Provide Context** - Share relevant background
3. **Break Down Complex Tasks** - Ask step by step
4. **Use Examples** - Show what you're looking for
### Effective Prompting
**Less Effective:**
Write code
**More Effective:**
Write a Python function that validates email addresses. It should:
- Check for @ symbol
- Verify domain format
- Return True/False
- Include docstring and type hints
### Working with Tools
1. **Enable Relevant Tools** - Only enable what you need
2. **Be Explicit** - "Search the web for..." helps AI know to use tools
3. **Review Results** - Tool outputs are included in context
## Troubleshooting
### Slow Responses
- Check your internet connection
- Gemini 1.5 Pro is slower but more thorough
- Very long conversations may slow down
### Unexpected Responses
- Rephrase your question
- Provide more context
- Start a new conversation for fresh context
### Tool Not Working
- Verify the tool is enabled
- Check MCP server connection status
- Ensure the tool matches your request
## Privacy & Data
- Conversations are stored locally in your browser
- Clear conversations to remove history
- API calls are processed by Google Gemini
- See our privacy policy for details
---
*Previous: [Studio](/docs/features/studio) | Next: [Agents](/docs/features/agents)*