API Integration
Also known as: External API, Third-Party Integration, Service Integration
Definition
Connecting an LLM application to external services via their APIs, enabling the model to read data from or trigger actions in other systems. API integration is how LLMs access the broader software ecosystem—weather services, databases, SaaS products, internal systems. It's the implementation layer beneath tool use.
What this is NOT
- Not the API itself (integration is connecting to it)
- Not tool use conceptually (integration is the implementation)
- Not retrieval (retrieval is a specific type of integration)
Alternative Interpretations
Different communities use this term differently:
llm-practitioners
Building tools that wrap external APIs so that LLM agents can invoke them. This includes handling authentication, request formatting, response parsing, error handling, and rate limiting.
Sources: LangChain API integration patterns, OpenAI plugins architecture (deprecated), MCP server implementations
software-engineering
The general practice of connecting software systems via their programmatic interfaces, handling authentication, data transformation, and error cases.
Sources: API design and integration literature
Examples
- A tool that calls the Slack API to send messages
- Integration with Stripe for payment processing
- Connecting to a company's internal REST API
- MCP server wrapping a SaaS product's API
Counterexamples
Things that might seem like API Integration but are not:
- The external API documentation (that's what you integrate with)
- The LLM deciding to use a tool (that's tool selection)
- Hard-coded logic that doesn't call external services
Relations
- requires tool-use (API integration implements tool use)
- overlapsWith tool-binding (Binding connects API integrations to tool definitions)
- overlapsWith model-context-protocol (MCP standardizes API integrations)
Implementations
Tools and frameworks that implement this concept:
- Doppler secondary
- n8n primary
- Playwright secondary