MENTIS

Modular Executive Neural Tooling Intelligence System

MENTIS is a local-first AI orchestration and developer tooling project focused on coordinating AI models, coding agents, project memory, command workflows, safety checks, and terminal-based execution.

This repository is the public project overview for MENTIS.

The full working repository is currently private while it is being cleaned, audited, and prepared for a staged open-source release. This public repository intentionally does not include private implementation code, local environment files, API keys, logs, personal memory, experimental artifacts, or unreleased internal tooling.

What MENTIS aims to become

MENTIS is designed as a transparent control layer for AI-assisted development.

Instead of relying on one opaque assistant, MENTIS aims to provide a structured workspace where builders can:

• route tasks to the right model or specialist agent
• coordinate coding, research, review, testing, and planning workflows
• manage project context and long-running development state
• keep humans in control of tool execution
• enforce safety policies before commands or file changes run
• build reliable terminal-first AI workflows
• compare local models and cloud models through one interface

Current status

MENTIS is in early private development.

The current focus is on:

• terminal user experience
• model routing
• agent-role orchestration
• Codex/OpenAI workflow integration
• safe command execution
• project memory boundaries
• public documentation and open-source release planning

Why this repository exists

This public repository provides a clean, non-sensitive overview of the project while the core implementation is prepared for release.

It exists to:

• explain the project publicly
• document the roadmap
• collect feedback
• prepare community contribution guidelines
• provide a stable link for programs, grants, and open-source review
• avoid exposing private or unsafe development artifacts too early

Planned open-source release

The planned public release will happen in stages:

1. Project overview and roadmap
   Public documentation, goals, architecture, contribution model, and safety policy.

2. Minimal developer preview
   A small, auditable CLI/TUI skeleton with no private keys, no personal memory, and no hidden automation.

3. Model routing layer
   Public implementation of provider abstraction, routing policies, and configurable model departments.

4. Codex/OpenAI workflow integration
   Reproducible examples for using Codex/OpenAI models in developer workflows.

5. Agent orchestration modules
   Specialist role coordination for coding, review, testing, research, and safety checks.

6. Community-ready package
   Documentation, examples, tests, security policy, and contributor workflow.

Non-goals

MENTIS is not intended to be:

• a malware framework
• an autonomous system that acts without user review
• a tool for credential theft, unauthorized access, or unsafe automation
• a black-box assistant that hides tool calls or model decisions
• a repository for private logs, secrets, or personal memory

Project principles

• Local-first where possible
• Human-in-the-loop by default
• Transparent model routing
• No secret leakage
• Safety gates before risky actions
• Useful terminal UX
• Practical developer workflows over hype
• Open-source release only after cleanup and audit

Repository scope

This repository currently contains documentation only.

The implementation is intentionally excluded until it can be released safely.

See:

• docs/ARCHITECTURE.md
• docs/ROADMAP.md
• docs/OPEN_SOURCE_RELEASE_PLAN.md
• docs/API_CREDITS_USE.md
• SECURITY.md

License

The documentation in this repository is released under the MIT License unless otherwise noted.

Future source-code licensing may be finalized before the first public implementation release.