Open Run

Give repeatable work a place to run

Open Run schedules coding-agent CLIs on a local machine. A project is a Git repository, an instruction is a prompt, and a trigger decides when the work starts.

It uses the agent CLIs a developer already has configured. Open Run itself does not require model API keys. The selected agent can still send prompts and code to its provider according to that agent’s settings.

A run should leave something you can review

The project includes schedules, one-off runs and webhook-triggered work, along with conversation history and Git diff review. Repeated work gets an execution directory tied to a recorded base commit.

A useful automation produces an understandable change. The review step connects the instruction, the result, the checks and the diff.

Examples of the kind of work it supports

A scheduled prompt might ask an agent to inspect a flaky test, explain a dependency change or prepare a small maintenance patch. These are example workflows, not promises that the tool will resolve every task unattended.

The machine and worker must be running for local schedules to fire. Agent availability and provider limits still apply.

Explore the implementation

The Open Run repository documents the CLI, runtimes, triggers and security model. It is the best place to check current behavior before configuring an automation.

Open Run sits beside cloudconfig and Nuxt Agentic in my work on making agent-driven applications and workflows easier to understand.