Devotd Github
NewAI agents

neopanic/harness — a composable Rust agent runtime

A Rust agent runtime that provides interfaces allowing the host to assemble the parts and boundaries needed for execution.

Repositoryneopanic/harness ↗

A Rust agent runtime that provides interfaces allowing the host to assemble the parts and boundaries needed for execution.

GitHub description: Composable Rust agent runtime with durable conversations, pluggable tools, skills, policies and context compaction

stars
★ 4
Forks
0
Language
Rust
License
MIT
Read original source ↗View on GitHub ↗

neopanic/harness is a composable agent runtime written in Rust. It does not include the workflow of any particular application; instead it provides interfaces that let the host choose and wire together the model and input, tools, prompts, policies, conversation store, observation capability, and transport boundaries. The execution loop can also be replaced. There is no published information about the latest release.

The purpose of this repository is not to provide the agent application itself, but to let the host assemble the parts and boundaries required for execution. It covers resume points, tool execution records, and policy mediation, along with the ability to preserve conversation state and continue from there. The ordering of parallel read calls is split out into its own convention. Tools, skills, policies, and context compaction can be swapped in, and it is published under the MIT license. So it is closer to a foundation for building an execution structure than to a single finished agent.

One caveat is that the repository's guidance distinguishes build and documentation generation checks from verification of actual runtime behavior. There are no tests in this repository that confirm runtime behavior; the repository explains that deterministic behavior cases belong to the applications that use it. The limits of external effects and checkpoint recovery are also included in the conventions, but they should not be read as material guaranteeing actual behavior. Therefore it is concrete as material for examining design boundaries, but runtime reliability is hard to judge from this repository alone.