OpenAI API for AI Agents

GPT models, embeddings, and AI infrastructure

OpenAI provides APIs for text generation, embeddings, image generation, and function calling. AI agents can use OpenAI as a reasoning engine, generate content, create embeddings for search, and leverage tool use capabilities.

What AI agents can do with OpenAI

Structured actions an AI agent can execute through the OpenAI API

Action

Description

Inputs

Outputs

createChatCompletion

Generate a chat completion response

model, messages, tools, temperature

message, tool_calls, usage

createEmbedding

Generate text embeddings

input, model

embeddings[], usage

createImage

Generate an image from a text prompt

prompt, model, size, quality

url, revised_prompt

listModels

List available models

models[], total

Use cases for OpenAI + AI agents

AI agent reasoning and decision making
Content generation and summarization
Semantic search via embeddings
Image generation from descriptions
Multi-step tool use with function calling

How to connect OpenAI to an AI agent

1Get your OpenAI API key
2Generate an AgentSpec on Elba for OpenAI actions
3Define completion, embedding, and image actions
4Publish for MCP config and discovery
5Set usage limits and rate controls

Best practices

✓Use function calling for structured outputs instead of parsing text

✓Set temperature based on task (0 for deterministic, 0.7+ for creative)

✓Implement token budget tracking to control costs

✓Use embeddings for retrieval-augmented generation

✓Cache frequent requests to reduce API calls

Frequently asked questions

How do AI agents use the OpenAI API?+

Agents call createChatCompletion with structured messages and tool definitions. The API returns either a text response or tool calls that the agent can execute. This enables multi-step reasoning and action.

What is OpenAI function calling?+

Function calling lets you define tools that GPT models can invoke. The model returns structured JSON matching your function schema instead of plain text. This is the foundation of agent tool use.

How do I control OpenAI API costs for agents?+

Set max_tokens limits, use cheaper models for simple tasks, cache embeddings, implement request budgets per session, and monitor usage through the OpenAI dashboard.