Tutor you through pen-and-paper math with a local agent
i open-sourced the app, although you can make your own quite trivially https://github.com/nichochar/arii borrowed the SOUL.md concept from openclaw, but repurposed as TUTOR.mdit uses codex locally so you can use your subscription credits rather than an API keyit's very fun! Task: The author open-sourced a local math tutoring agent named 'arii' that uses Codex with subscription credits instead of API keys. The workflow involves creating a TUTOR.md file, adapted from the SOUL.md concept, to define the agent's role for pen-and-paper math assistance. The tool sequence relies on local execution of Codex to interact with the user, though specific human oversight steps are not detailed in the source.; tools: Codex. Human role: The original post does not say how a person stayed involved.; limitations: Prerequisites include access to a local Codex environment and a valid subscription for credit usage. The source does not specify integration constraints, operating costs beyond subscription fees, or exception handling mechanisms. Reliability is secondary as the claim comes from a single user post without independent verification. Human approval processes are not defined in the original text..
Tools
Published on Agent Work Atlas:
Overview
i open-sourced the app, although you can make your own quite trivially https://github.com/nichochar/arii borrowed the SOUL.md concept from openclaw, but repurposed as TUTOR.mdit uses codex locally so you can use your subscription credits rather than an API keyit's very fun!
Workflow
The author open-sourced a local math tutoring agent named 'arii' that uses Codex with subscription credits instead of API keys. The workflow involves creating a TUTOR.md file, adapted from the SOUL.md concept, to define the agent's role for pen-and-paper math assistance. The tool sequence relies on local execution of Codex to interact with the user, though specific human oversight steps are not detailed in the source.
Reported result
The source does not report any measured outcomes, performance metrics, or specific business results from using this agent. The author describes the experience as 'very fun' but provides no data on effectiveness or user retention. A measurable acceptance check would require defining success criteria such as problem-solving accuracy or user satisfaction scores, which are currently unknown.
Human role
The original post does not say how a person stayed involved.
Limitations
Prerequisites include access to a local Codex environment and a valid subscription for credit usage. The source does not specify integration constraints, operating costs beyond subscription fees, or exception handling mechanisms. Reliability is secondary as the claim comes from a single user post without independent verification. Human approval processes are not defined in the original text.
Try it
I want to set up a local math tutoring agent using Codex and a TUTOR.md configuration file. Please guide me through cloning the 'arii' repository from https://github.com/nichochar/arii or creating a similar structure. First, help me draft a TUTOR.md file that defines the agent's persona for pen-and-paper math assistance, inspired by the SOUL.md concept. Ensure the setup uses my local Codex subscription credits rather than an API key. Before running any full sessions, perform a dry run with one sample math problem to verify the agent's response quality. Show me the output and wait for my confirmation before proceeding to further interactions. If any configuration files are missing, ask me for the specific details needed to complete the setup.
Source reports are not independently verified. Publication date is the catalog publication date, not the original article date.