llmcloud.ai
Get API key
Apps · Autonomous agents

CrewAI

Role-based multi-agent orchestration framework.

Tokens / 7d
52B
Routing directive
auto:agent
On our metal
29%
Licence
Open source

CrewAI crews mix an expensive planner with several cheap workers. That shape is a natural fit for per-agent routing: the manager gets a reasoner, the workers get a fast small model.

→ crewai.com→ All apps→ Integration docs

How it connects

  • LiteLLM-style model strings resolved through the gateway.
  • Manager agent on auto:reasoning, worker agents on claude-haiku-4.5.
  • Shared cache across agents removes duplicated context re-reads.

Traffic pattern

Fan-out. One planning call spawns 5–20 worker calls, most of which share a large common prefix.

gpt-5.5claude-haiku-4.5qwen3-max

Point it at the gateway

from crewai import Agent, LLM
manager = Agent(llm=LLM(model="auto:reasoning",
  base_url="https://api.llmcloud.ai/v1"))
worker  = Agent(llm=LLM(model="claude-haiku-4.5",
  base_url="https://api.llmcloud.ai/v1"))