Runprompt is a tool for executing LLM prompts from the command line. It was developed with reference to Google's Dotprompt format.
Runprompt Released: LLM Prompting via Unix Pipes in a Single-File Python Script
This article is a translation. Read the Japanese original
This format consists of front matter and Handlebars templates. The developer stated that they wanted to treat prompts as programs that could be connected via Unix-style pipes.
Specifically, structured output is enabled by defining a JSON schema. The JSON returned by the LLM can then be piped into other tools.
Additionally, it features the ability to take the output of one prompt as a template variable for the next. This allows for the construction of multi-step agent workflows using shell pipelines.
The only dependencies are Python's standard libraries. It operates simply by downloading and running it via curl.
Supported providers include Anthropic, OpenAI, Google AI, and OpenRouter. Access to dozens of models is available via OpenRouter.
It can be used for automation without a framework, such as extracting data from unstructured text or generating reports from logs.
Source: Show HN: Runprompt – run .prompt files from the command line (HN 134pt, 49 comments) (HN Search (backfill), 2025-11-27)