> ## Documentation Index
> Fetch the complete documentation index at: https://praison.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# AI Agents
> Overview of all available PraisonAI agents and their capabilities
Every specialised agent starts from the same friendly `Agent` primitive — one instruction, one call — then adds tools or dedicated methods for its domain.
```python theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
from praisonaiagents import Agent
agent = Agent(instructions="Your task-specific instructions")
agent.start("Your task description")
```
```mermaid theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
graph LR
subgraph "PraisonAI Agents"
Task[📋 Task] --> Agent[🤖 AI Agent]
Agent --> Tools[🧠 Tools / SDK]
Tools --> Output[✅ Result]
end
classDef input fill:#6366F1,stroke:#7C90A0,color:#fff
classDef process fill:#189AB4,stroke:#7C90A0,color:#fff
classDef output fill:#10B981,stroke:#7C90A0,color:#fff
class Task,Agent input
class Tools process
class Output output
```
PraisonAI provides a diverse set of specialized agents for various tasks. Each agent is designed with specific capabilities and tools to handle different types of tasks effectively.
## Quick Start
Create a plain agent and start it.
```python theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
from praisonaiagents import Agent
agent = Agent(instructions="Your task-specific instructions")
agent.start("Your task description")
```
Swap in a dedicated agent class when a task needs it.
```python theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
from praisonaiagents import CodeAgent
agent = CodeAgent(name="Coder")
code = agent.generate("Write a function to reverse a string")
print(agent.execute(code)["stdout"])
```
## How It Works
```mermaid theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
sequenceDiagram
participant User
participant Agent
participant Tools as Tools / SDK
User->>Agent: Task description
Agent->>Tools: Invoke the right tool or method
Tools-->>Agent: Result
Agent-->>User: Final response
```
## Data & Analysis
Analyze data from various sources, create visualizations, and generate insights.
Track stocks, analyze financial data, and provide investment recommendations.
Conduct comprehensive research and analysis across various topics.
Search and extract information from Wikipedia articles.
## Media & Content
Analyze and understand visual content from images.
Convert images to textual descriptions and extract text content.
Analyze video content and extract meaningful information.
Generate and format content in Markdown syntax.
## Search & Recommendations
Perform intelligent web searches and gather information.
Generate personalized recommendations based on preferences.
Compare prices and find the best deals across stores.
Create travel plans and detailed itineraries.
## Development
Write, analyze, and debug code across multiple languages.
Simple, focused agent for basic tasks without external tools.
## Getting Started
Each agent can be easily initialized and customized for your specific needs. Here's a basic example:
```python theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
from praisonaiagents import Agent
agent = Agent(instructions="Your task-specific instructions")
response = agent.start("Your task description")
```
For more detailed information about each agent, click on the respective cards above.
## Best Practices
A plain `Agent` with a clear instruction handles most tasks. Reach for a specialised class only when you need its dedicated methods or built-in tooling.
Use `CodeAgent` for sandboxed code, `VisionAgent` for image understanding, `RealtimeAgent` for voice, and so on. Each ships the right defaults for its domain.
Many "agents" here are just a plain Agent plus a tool (search, finance, wiki). Attaching a tool is often simpler than adopting a whole specialised class.
Set `memory=True` when a task spans several interactions so the agent keeps context across turns instead of starting fresh each time.
## Related
The minimal single-purpose agent to start from.
The CodeAgent with sandboxed generation and execution.