QuantMD

Isn’t it just crazy that one needs so much programming expertise just to write the code to orchestrate two AI agents, using existing libraries?

CrewAI: 15,000 lines of code.
LangGraph: 50,000 lines.
The thing I actually needed: ~150 lines.

So I built litecrew.

Here's a complete multi-agent workflow:

researcher = Agent("researcher", model="gpt-4o-mini")
writer = Agent("writer", model="claude-3-5-sonnet")

(researcher, writer)
def write_article(topic):
research = researcher(f"Research {topic}")
return writer(f"Write about: {research}")

That's it. No YAML. No config files. No framework to learn.

What it does:
→ Sequential handoffs (A → B → C)
→ Parallel ex*****on (fan out, collect)
→ Tool calling
→ Token tracking
→ Works with OpenAI, Anthropic, or Ollama

What it doesn't do (on purpose):
→ No hierarchical agent management
→ No state machines
→ No streaming

The philosophy: SQLite doesn't try to be PostgreSQL. Sometimes simple is the feature.

If you need the full power of CrewAI, use CrewAI. This is for when you just want agents to pass data without reading 200 pages of docs.

Open source. MIT licensed. Your API keys stay on your machine.

Link in comments 👇

A little interactive app from QuantMD to offer a sense for daily deaths and cumulative deaths owing to COVID-19:
https://quantmd.shinyapps.io/visualizeCorona/

Data is updated daily (last update 3/14)