heymrun/heym is an environment for deploying, running and observing agent systems on your own infrastructure. The repository's latest release is v0.0.138. The provided information contains no changes for this version, so no newly added or modified features can be identified. According to the README, workflows can be built and run with a visual editor or configuration files. It orchestrates agents together with defined task steps, tools, parallel branches, retries and human approval steps.
The project is built not only to automate an agent's work but also to let you watch the execution itself. The README explains that run history, tool calls, latency, errors and model cost can be checked and the results evaluated. It also presents ways to expose capabilities through REST, SSE and MCP interfaces and an interactive portal. It is written in Python, and the repository description lists business process automation, execution review and human control as its main uses.
However, the provided information contains no performance comparisons or benchmark figures, and the license is recorded as NOASSERTION. The momentum score in the evaluation data is 3.4, and no specific changes for the new release were given. The scope as an execution environment with self-hosting and approval procedures is therefore confirmed, but the basis for judging performance, license conditions or the extent of improvement in this version is limited.