Tool Binding

Process tools published

Also known as: Tool Registration, Tool Configuration

Definition

The process of connecting tool definitions to their implementations and making them available to an LLM or agent. Binding involves: defining the tool schema, implementing the execution logic, registering the tool with the application, and configuring any authentication or permissions. It's the bridge between "here's what this tool does" and "here's how to actually run it."

What this is NOT

  • Not the tool definition alone (binding connects definition to implementation)
  • Not tool execution (binding sets up tools; execution runs them)
  • Not the model's decision to use a tool (that's tool selection)

Alternative Interpretations

Different communities use this term differently:

llm-practitioners

Configuring an agent framework (LangChain, LlamaIndex, etc.) with the tools it can use, including both the definitions sent to the model and the Python/JS functions that execute when tools are called.

Sources: LangChain Tool documentation, LlamaIndex tool usage, Agent framework documentation

Examples

  • LangChain @tool decorator binding a Python function as a tool
  • Registering MCP server tools with Claude Desktop
  • Configuring OpenAI Assistants with file_search and code_interpreter tools
  • Binding a database query function with authentication credentials

Counterexamples

Things that might seem like Tool Binding but are not:

  • The tool definition in isolation (that's schema, not binding)
  • The model outputting a tool call (that's invocation)
  • The user deciding which tools to expose (that's configuration, pre-binding)

Relations

See Also

Implementations

Tools and frameworks that implement this concept: