Show HN: Agent Workflow Language – Open-Source Scala DSL for LLM Orchestration
awfl.usGitHub: https://github.com/awfl-us
AWFL is an open source suite for building and orchestrating LLM-driven agents. It includes:
- A simple Scala DSL for composing workflows that compile to Google Workflows - Node services providing context, tools, and task orchestration for agents - A customizable web UI for managing projects, sessions, and agents - A CLI tool for granting agents local file access (pipx install awfl)
Why: AWFL is meant to be a transparent, locally runnable foundation for anyone building multi-agent or workflow-based AI systems.
Examples / POC: https://github.com/cornerstone-ai https://www.cornerstoneai.org/
Hello, quick note about trying out awfl.us:
If you create a project and start a session without running awfl locally, it will spin up a cloud consumer. This takes a little time, but it provides a safe, isolated environment where your agents can run commands and make file updates.
My main motivation for building the Scala DSL was to have a clean, functional orchestration core. The Node services then extend functionality by handling the more procedural logic that LLMs are better at implementing.
Feedback on the API design, DSL ergonomics, and overall architecture would be much appreciated.