Runie — We're going straight forward to the future where you will manage bunch of models and agents
We're going straight forward to the future where you will manage bunch of models and agents. Why to not make a harness with a native support of it?
What Runie Does
Runie registers any LLM model you use — OpenAI, Anthropic, local models, and more. It orchestrates multi-agent swarms with Erlang-style supervision, measures if your agent is actually improving, and ships Rust code. MIT licensed.
Key Features
Rust-native coding agent harness with terminal UI (TUI)
Register any LLM model — OpenAI, Anthropic, local, 25+ providers
Orchestrates agent swarms with Erlang-style supervision
Ships actual Rust code, not just chat
Actually measures if your agent is improving
Pattern-aware skills system
Multi-agent orchestration
MIT licensed
Model-Agnostic
Runie works with any LLM provider. Register OpenAI models, Anthropic models, local models, or any of the 25+ providers supported. Swap models with a single environment variable. No vendor lock-in.
Erlang-Style Supervision
Agent swarms are orchestrated with Erlang-style supervision patterns. Fault tolerance, rest-for-one strategies, and hierarchical supervision trees keep your agents running reliably.
Actually Measure Improvement
Stop guessing if your agent improved. Runie actually measures it. Built-in evaluation lets you track performance over iterations and ship with confidence.