Multi-Tool Reasoning
Learn how to build coding assistants that use multiple tools across turns with the Autohand Code Agent SDK. Available in TypeScript, Python, and Java.
In this tutorial, you'll build a coding assistant that uses multiple tools across different turns to understand code, run tests, and report a summary. This demonstrates the ReAct loop with multi-tool turns.
- Completed Code Reviewer Agent
- Node.js 16+ (for TypeScript), Python 3.10+ (for Python), or Java 11+ (for Java)
- SDK installed for your chosen language
- API key from your LLM provider (set as
AUTOHAND_API_KEY)
What You'll Build
A code analyst coding assistant that:
- Uses READ_FILE to examine source code
- Uses BASH to run commands and tests
- Uses GLOB to find files
- Coordinates multiple tools across turns
- Provides a comprehensive codebase summary
Step 1: Direct Tool Execution
Before building the agent, let's understand direct tool execution in your chosen language:
Step 2: Create the Multi-Tool Agent
Now create an agent that uses multiple tools across turns in your chosen language:
Step 3: Run the Agent
Run the agent with a complex prompt that requires multiple tools in your chosen language:
Key Concepts
The coding assistant automatically coordinates multiple tools across turns to complete complex tasks.
The ReAct (Reason + Act) loop allows the coding assistant to reason about what to do, take action, and repeat.
The coding assistant maintains context across turns, remembering what it has done and what it needs to do next.
Setting maxTurns prevents infinite loops and controls resource usage.
Complete Example
Here's the complete example with a sample project:
Next Steps
🔧 Sequential Agent Pipeline
Learn how to chain multiple agents together for complex workflows.
View Example →