Practical overview
When it fits
Use it when you want to author agent logic as Markdown and operate it locally through a CLI rather than write Python scaffolding for the workflow.
Who it is for
- Developers who want to express agent instructions and control flow in Markdown files.
- People building command-line agent workflows that use tools, memory or linked instructions.
- Experimenters who want to assemble or connect their own reasoning approaches.
Documented uses
- Define an agent’s instructions and their references in a .gen file.
- Run an agent through the command-line interface.
- Build example content workflows for collecting video information, transcribing videos or drafting topic-based articles.
Requirements & limits
The project is a public GitHub repository under the MIT license. Its documented setup requires an OpenAI API key.
- The README documents OpenAI configuration and notes Llama 3 support through providers compatible with the OpenAI API.
- Building requires the listed C++ dependencies and Make; the README provides dependency instructions for Linux, macOS and Windows.
- The README marks the local vector database and web UI as work in progress.
- The README says Python can be integrated into Markdown agent scripts when needed.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Build Mentals AI and run the included loop agent
Follow the README’s documented setup path to build the CLI and invoke its example agent.
Before you start
- An OpenAI API key
- libcurl, libfmt and poppler dependencies
- A terminal and Make
Steps
Install the listed dependencies using the README’s instructions for your operating system.
Clone the repository and enter its directory.
Add the API key, endpoint and model settings to config.toml.
Run make, then invoke ./build/mentals agents/loop.gen -d.
Expected outcome
The documented command invokes Mentals AI with the agents/loop.gen example file.
Guide limitations
- The README documents setup and invocation but does not specify the example’s output.
- The README says pgvector installation can be skipped on the main branch.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 14 October 2026
- Original website title
- turing-machines/mentals-ai · Official homepage title · retained in its source language

