Skip to content

Getting Started

G6 is a tool server that connects to your AI assistant. After setup, your assistant gains access to 262 MCP tools — formal proofs, ML pipelines, safety checks — that it can invoke as naturally as its built-in capabilities.

What to expect

  1. Connect G6 to your assistant (a few minutes) — create an account, get an API key, and point Claude Code (or any MCP client) at the hosted server. Nothing to download or install
  2. Make your first tool call — ask G6 (or your assistant) to decompose a goal or verify a formula
  3. Explore capabilities — browse tools by use case and build them into your workflows

Choose your path

  • Setup

    Download G6, activate your license, and connect in a few minutes.

    Setup

  • Choose Your Interface

    GUI, terminal (TUI/CLI), Claude Code (MCP), or REST — and what runs locally.

    Choose Your Interface

  • Building on G6

    Build extension components on top of the installed app with the g6ext.* SDK.

    Developer Guide

Quick starts by role

  • Business Operator

    Automate operations using your existing data without AI expertise. Start with the GoalWizard for natural language goals.

    Business quick start

  • Domain Worker

    Get AI tools purpose-built for your profession. 39 job agent specialisations from accounting to cybersecurity.

    Worker quick start

  • Software Engineer

    Extend your coding agent with composable AI. GUI, MCP, terminal, and REST surfaces.

    Choose Your Interface

  • Domain Specialist

    Build reliable applications in your field with self-training loops, safety gates, and optional formal verification backends.

    Specialist quick start

What changes after setup

Once G6 is connected, your AI assistant can:

  • Reason formally — prove properties, check satisfiability, and verify specifications instead of guessing
  • Learn from data — run ML pipelines, optimise parameters, and persist what it learns across sessions
  • Work safely — every tool call passes through the Constructive Safety Framework, which bounds risk before execution rather than catching failures after

For the full value proposition, see Architecture Overview.

Your learning path

  1. Configuration — environment variables, CSF settings, and the Settings class
  2. Troubleshooting — solutions for common errors on Windows, Linux, and macOS
  3. Use Cases — practical workflows for code review, research, and data analysis
  4. Architecture — how the system is built and why