Agents, Reliability & Cost · Intermediate
ReAct Agent Playground
Watch an agent think, call a tool, observe the result and repeat until it reaches an answer.
Reasoning trace
Tool calls
Observations
Understanding ReAct Agents
Reasoning (Thought)
The agent thinks through the problem, analyzing what it knows and what it needs to find out.
Breaking down complex tasks into steps
Example: "I need to search for the current weather"
Acting (Action)
The agent takes actions using available tools to gather information or make changes.
Using tools like search, calculator, APIs
Action: Search["weather in San Francisco"]
Observation (Result)
The agent receives feedback from actions and incorporates it into reasoning.
Learning from tool outputs
Observation: "Sunny, 72°F in San Francisco"
ReAct Loop
1
Thought: Analyze the problem
→
2
Action: Use a tool
→
3
Observation: Get result
→
4
Repeat: Until solved
Agent Configuration
Available Tools:
Search
Calculator
Weather API
Database
Calendar
Email
Steps Taken
0
Tools Used
0
Status
Ready
Execution Trace
Select a task and click "Run Agent" to see the reasoning process