AutoMates

Where Your Imagination
Becomes Creation

Clone a folder. Your AI starts remembering between sessions, researching before it codes, and reviewing its own work.
No app, no account — just markdown files that restructure how your AI thinks.

Persistent memory. Professional knowledge. Built-in review.
You're the pilot — they're the crew.
Specialists that research, build, and check each other's work.

How It Works

From zero to a full AI dev team in 5 minutes.

0

What You Need

One of these AI tools installed:

Each requires a subscription or API key — check their pricing pages.

Also works inside VS Code and JetBrains. Won't touch your existing config.

1

Download and Open

In your terminal:

git clone https://github.com/Nate-Vish/Auto-Mates.git cd Auto-Mates

Then start your AI tool by typing claude, gemini, or codex.

Your AI reads the files in this folder and uses them as instructions, memory, and workspace — turning one AI into a coordinated team of specialists.

2

Meet Your Team Lead

Orca (the orchestrator) loads automatically — it reads the project, your past sessions, and walks you through everything. No manual required.

First time? Just say "Give me a tour."

3

Tell It What You Want

Describe what you want to build. Orca picks the right specialists — Planner designs it, Builder codes it, Checker reviews it.

You are the pilot. They are the crew.

4

Save Your Progress

When you're done, type /memorize in the chat. Your team saves everything — what happened, what worked, and your preferences. All saved as text files in your local folder — no cloud, no account needed.

5

Come Back Anytime

Close the terminal. Next time you start up, your team picks up right where you left off — what you built, where you stopped, and what's ahead.

Built Different

Drop-In Ready

Won't break your existing setup. Uses .claude/rules/ instead of overwriting your config. Clone it into any project — it just works alongside what you already have.

Learn First

Agents research before they code. Fetcher gathers, indexes, and verifies sources locally — engineered to filter out misleading data. Like having a research department, not a search bar.

Transparent

Every agent identity is a readable file. Every decision is traceable in plain markdown. No black boxes, no hidden prompts.

Real Knowledge

Professional topic files across every agent. Not just role descriptions — actual expertise. Agents arrive with real craft on day one.

?!

Built-In Skepticism

Gal exists to question everything. The skeptical senior dev who catches what you missed. AI that reviews AI — before it ships.

Your Chief of Staff

Orca loads on day one, knows every agent and every folder, remembers your last session, and routes you to exactly who you need. Like walking into a company where someone already knows your name.

The Structure

AgenTeam
AgenTeam
Agent identities, memory, and session history
Orca
Orca
Orchestrator — routes tasks, leads the team
Planner
Planner
Architect — blueprints and project plans
Builder
Builder
Engineer — writes production code
Checker
Checker
Auditor — security review and quality gates
BrainStorm
BrainStorm
Knowledge Graph — ideas and connections
Legal
Legal
Compliance — licenses, privacy, governance
GitDude
GitDude
Release Manager — version control, releases
Gal
Gal
User Advocate — skeptical senior dev
Daisy
Daisy
Brand Director — branding, copy, pitches
Fetcher
Fetcher
Researcher — gathers docs, sources, intelligence
Library
Library
Knowledge base, research sources, and agent routing
Knowledge/
Per-agent professional expertise — curated topic files
Sources
Sources/
Research library — agents study here before working
Registry.md
Agent routing truth — who handles what
Rules.md
Project constraints and guardrails
Arsenal/
Tools, MCPs, and integrations — shared across all agents

Built in the Open

AutoMates is open source. Star the repo, fork it, forge your own agents. The best tools are the ones you can shape.

GitHub