It is a "big one". Theres too much to explain in a few words.

- The by a large distance most extensive possibilty to define Agents, Teams and their message connections via YAML.

- Ability to consruct Agent workflows by two different approaches, either by creating a hierarchy of Agents, parallel teams or sequential teams, or by "connecting" Agents and creating observable and even routing-capable Connection objects

- Support for both Pydantic-AI and LiteLLM as LLM backends

- MCP server support for agents

- Strong focus on pydantic-level Typing (Structured responses, generically typed communication objects)

- Nested Agent teams which can be used like "regular" agents since both teams and Agents share a same base interface.

- A CLI which allows to create Agent workflows using a slash-command-system

and MUCH MUCH more. its really to much to cover. It doesnt have that many integrations yet, but apart from it, this kicks AutoGens and CrewAIs a*. ;)

Check out

https://phil65.github.io/llmling-agent

https://github.com/phil65/llmling-agent

Cheers and thank you!