For AI agents

A free LLM backend for your agents

Autonomous agents make a lot of model calls. DreamPrompting gives them a free, OpenAI-compatible endpoint with automatic failover across 27 providers, so a single provider's rate limit does not stop your agent loop.

Read the API docs Building a coding agent? Try the chat

Drop-in compatible

Anything that speaks the OpenAI Chat Completions API works. Change the base URL, keep your agent code.

Failover = uptime

If a provider is down or rate-limited, the gateway retries the next one. Your agent does not crash on a single 429.

One free key

Grab a free key and go. 100 requests/minute, no credit card.

Wire it into your framework

LangChain (Python)
from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    base_url="https://dreamprompting.com/api/v1",
    api_key="YOUR_API_KEY",
    model="auto",
)
print(llm.invoke("Plan a 3-step task").content)
CrewAI
from crewai import LLM

llm = LLM(
    model="openai/auto",
    base_url="https://dreamprompting.com/api/v1",
    api_key="YOUR_API_KEY",
)
AutoGen
config_list = [{
    "model": "auto",
    "base_url": "https://dreamprompting.com/api/v1",
    "api_key": "YOUR_API_KEY",
}]
LlamaIndex
from llama_index.llms.openai_like import OpenAILike

llm = OpenAILike(
    model="auto",
    api_base="https://dreamprompting.com/api/v1",
    api_key="YOUR_API_KEY",
    is_chat_model=True,
)

Agent-builder notes

  • Respect the rate limit (100/min per IP). For tight agent loops, add small delays or backoff on 429.
  • Pin a model for consistency (e.g. groq/openai/gpt-oss-120b) or use auto for maximum resilience.
  • Streaming is supported for responsive agent UIs (stream: true).
  • Free with your API key. Need a higher limit for heavy production loads? Just ask, or run your own upstream provider keys directly.