⚡ OpenGuild (antigravity-guild)

Turn your AI coding assistant into a 16-person staff engineering team that never forgets a bug, never leaks a secret, and saves thousands of tokens on every chat turn.

Run one command in any project:

npx antigravity-guild

(or npx openguild)

Works out-of-the-box with Antigravity IDE, Cursor, Claude Desktop, VS Code, Windsurf, and Gemini.

😫 The Problem with AI Assistants Today

If you use Cursor, Claude, or Antigravity to write code every day, you know these frustrations:

1. AI Amnesia: You spend 30 minutes teaching your AI not to use any in TypeScript or how your auth middleware works. The next day, in a new chat, it makes the exact same mistake.
2. Context Bloat & Sluggish Responses: Huge system prompts burn 4,000+ tokens before you even type your prompt. Responses feel slow, cost real API dollars, and the AI suffers from attention drift.
3. Shallow, Junior Code: Most AI code works for a demo, but collapses in production: missing null checks, zero error handling, unparameterized SQL queries, and sloppy architecture.
4. Zero Team Alignment: Everyone on your team prompts their AI differently, resulting in mismatched styles, broken conventions, and conflicting PRs.

✨ How OpenGuild Fixes This

| Without OpenGuild | With OpenGuild v2.6.0 | |:---|:---| | AI forgets project lessons every time you open a new chat | Cross-project memory vault remembers every architectural rule and bugfix forever | | 4,500+ tokens burned on monolithic rules per turn | Lean 770-token base contract (80-85% token diet) with instant sub-second streaming | | AI generates naive code without thinking about edge cases | 16-Mind Council applies Google scalability, Apple UX aesthetics, and Anthropic safety | | AI commits broken code and forgets tests | Deterministic Verification Gate blocks completion until tests and linters pass | | Secrets, API keys, and raw eval() leak into code | Built-in SAIF 2.0 Security Auditor catches credentials and hazards before you commit |

⚡ Quick Start: Zero Friction, "It Just Works"

Like an Apple product, OpenGuild eliminates configuration rituals. You don't need multiple commands, complex flags, or magic prompting rituals.

Navigate to any project on your computer and run one single command:

cd your-project
npx antigravity-guild

🍎 What OpenGuild Does Automatically in 1 Second (99.99% Done):

1. Links Your Global Persistent Memory Hub (~/.openguild/memory): Connects past lessons, bug fixes, and architectural standards learned across all previous projects.
2. Initializes Workspace Team Memory (.openguild/): Sets up version-controlled project memory so teammates automatically share project-specific architecture decisions.
3. Auto-Detects Universal Stack: Identifies language (JS/TS/Python/Go/Rust/Flutter/Swift), framework, package manager, and monorepo tooling.
4. Auto-Configures Editor MCP (Zero-Click): Configures OpenGuild's native MCP server in Cursor, Claude Desktop, Antigravity / Gemini, VS Code, and Windsurf.
5. Synthesizes AI Agent Rules (Zero-Prompt Mandate): Writes AGENTS.md, .cursorrules, and .gemini/rules.md so the AI autonomously reviews memory, runs tests, and applies council rules without special prompting.
6. Enforces Context Hygiene: Updates .gitignore to protect API keys, credentials, and exclude AI context noise.
7. Auto-Lints & Cleans Memory Vault: Deduplicates and prunes redundant rules to keep your persistent memory razor-sharp.
8. Runs SAIF 2.0 Security Audit: Scans the codebase for committed secrets, dangerous code execution, and SQL hazards, assigning an instant letter grade (A+ to F).
9. Profiles Context Diet & Token Savings: Estimates active token load vs monolithic prompts, showing 80-85% token reduction and dollar savings.
10. Installs Git Pre-Commit Hooks & Generates CI: Automatically installs deterministic invariant pre-commit hooks and generates GitHub Actions council workflows.

💬 Just Chat Normally — Zero Prompt Burden

You never need to remember to tell your AI: "Check memory" or "Act as the 16-mind council".

Even if you type a short, casual prompt:

"fix the checkout bug"

Your AI assistant will autonomously:

1. Search your global memory for past payment/checkout lessons.
2. Review the code through Architecture, Security (SAIF 2.0), and QA lenses.
3. Fix the issue, run the project test suite, and record any new lessons so future projects never make the same mistake.

🛠️ The 6 Core Superpowers

1. 🧠 Persistent Memory That Actually Learns

Never repeat yourself to an AI again. Whenever you solve a tricky bug or decide on an architectural rule, teach OpenGuild:

# Save an insight to your global memory (available across all your projects)
npx antigravity-guild --learn "Always set Content-Type header on custom fetch calls" --category architecture

# Or teach it directly inside your chat with Cursor or Claude:
# "Remember: We always use zod to parse inbound request bodies in this project."

OpenGuild automatically scrubs API keys and secrets before saving, indexes the lesson with tags, and deduplicates redundant rules.

2. ⚡ Context Diet & Token Profiler (--tokens)

Monolithic prompts slow down your AI and cost money. OpenGuild keeps base instructions razor-thin (~770 tokens) and pulls deep knowledge only when needed via MCP tools.

Run the profiler anytime to see your token weight and savings:

npx antigravity-guild --tokens

Example Output:

# ⚡ OpenGuild Token & Context Diet Report
Efficiency Grade: A+ (🟢 Highly Optimized)
  Active Context Token Load:  771 tokens
  Standard Monolithic Prompt: 4,500 tokens
  Context Diet Savings:       83% reduction (~3,729 tokens saved/turn)
  Estimated Cost Savings:     ~$11.19 per 1,000 chat turns

3. 🔮 Autonomous Auto-Analyst (--analyze)

Have a rough idea but aren't sure how to architect it? The Auto-Analyst turns raw concepts into full production-grade blueprints in seconds:

npx antigravity-guild --analyze "A self-hosted bookmark manager with AI auto-tagging and offline search"

What it generates:

• Inferred problem domain and target personas
• Latent edge cases and failure modes you didn't think of
• Recommended tech stack (Frontend, Backend, Database, AI models)
• Apple-grade UI/UX specifications and micro-interactions
• Security guardrails and deterministic test contracts

4. 🧪 Deterministic Verification Gate (--verify)

Stop trusting an AI that says "Everything is done and working!" when tests are failing. Run OpenGuild's unified invariant gate:

# Verify tests, linter, types, memory integrity, and git hygiene in one command
npx antigravity-guild --verify

# Auto-fix memory duplicates and missing .gitignore rules
npx antigravity-guild --verify --fix

It executes your project's native test commands (npm test, pytest, cargo test, go test) and outputs a certified invariant proof.

5. 🛡️ Deep Security & SAIF 2.0 Codebase Auditor (--audit)

Scan your repository for secret leaks, hardcoded credentials, and high-risk code patterns before pushing to production:

npx antigravity-guild --audit

What it catches:

• Plaintext API keys, AWS credentials, GitHub tokens, and private keys
• Uncommitted sensitive .env files
• Dynamic SQL string concatenation (SQL injection hazards)
• Arbitrary code execution hazards (eval())

Assigns an institutional grade (A+ to F) with exact filenames, line numbers, and fix instructions.

6. 👥 Git-Shared Team Memory (--team)

Ensure every developer on your team gets the exact same high-quality AI output:

npx antigravity-guild --team

Creates a version-controlled .openguild/ folder containing team_memory.md and architecture_decisions.md. When teammates pull the repo, their AI assistants automatically inherit all team decisions.

💡 The Prompt Playbook (Copy & Paste Into Your Chat)

Once OpenGuild is set up, you don't need complex prompting. Copy and paste these battle-tested prompts into Cursor, Claude, or Antigravity:

🚀 1. Before Starting a New Feature (System Design)

"Act as the Chief Software Architect and 16-Mind Council. Review this feature request: [Describe Feature]. Design the data model, API contracts, edge cases, and component boundaries before writing any code."

🎨 2. Polishing UI & User Experience (Apple Standard)

"Review this screen through the lens of the Apple CTO and UI/UX Lead. Make typography, responsive layout, spacing, accessible contrast, and micro-interactions feel clean, premium, and intuitive."

⚡ 3. Optimizing Slow Code or Database Queries

"Have the Performance Specialist and Database Reliability Engineer review our queries, caching, and state management. Identify any N+1 query patterns, memory leaks, or unindexed lookups."

🛡️ 4. Security & Bug Audit (Before Submitting PR)

"Act as the Chief Security Officer and QA Lead. Audit this code for unhandled null/undefined values, secret leakage, unparameterized queries, and race conditions."

🧪 5. Deterministic Verification Gate

"Run our automated test suite and typechecks using the openguild_verify_invariants tool. Do not mark this task as complete until all tests pass deterministically."

🏛️ The 16-Mind Supreme Council

OpenGuild models its engineering guidelines after 16 specialized roles. Each mind focuses on a critical pillar of production software:

graph TD
    User([Developer / User]) --> OpenGuild[OpenGuild Unified Brain]
    OpenGuild --> Core[Core Engineering]
    OpenGuild --> BigTech[Big Tech CTO Minds]
    OpenGuild --> Specialists[Deep Domain Specialists]
    OpenGuild --> AGI[Autonomous & AGI Systems]

    Core --> Arch["🧠 Chief Software Architect"]
    Core --> UI["🎨 Principal UI/UX Lead"]
    Core --> Sec["🛡️ Chief Security Officer"]
    Core --> Eng["⚡ Staff Full-Stack Engineer"]
    Core --> QA["🧪 Principal QA Lead"]
    Core --> DevOps["🚀 DevOps Engineer"]

    BigTech --> Google["🌐 Google CTO Brain (Scalability)"]
    BigTech --> Apple["🍎 Apple CTO Brain (Aesthetics)"]
    BigTech --> Anthropic["🧭 Anthropic Safety Brain (Alignment)"]

    Specialists --> Perf["⚡ Performance Specialist (O(n), Caching)"]
    Specialists --> DB["🗄️ Database Reliability Engineer (ACID)"]
    Specialists --> Prod["💼 Product Strategy Lead (Anti-Bloat)"]

    AGI --> Cog["🔮 Chief Cognitive Analyst (Latent Edge Cases)"]
    AGI --> Sci["🔬 Principal AGI Scientist (Cognitive Loops)"]
    AGI --> Heal["🧬 Self-Healing Specialist (Fault Recovery)"]
    AGI --> Graph["📊 Knowledge Graph Architect (Vector & RAG)"]

🌐 Model Context Protocol (MCP) Server

OpenGuild exposes 10 native MCP tools that give AI assistants live superpowers inside your editor:

| MCP Tool | What It Does for the AI | |:---|:---| | openguild_auto_analyze | Auto-analyzes product visions and codebases to generate complete engineering blueprints | | openguild_profile_tokens | Checks AI rule token weight, efficiency scores, and context savings | | openguild_verify_invariants | Runs hermetic verification across tests, linters, types, memory, and git hygiene | | openguild_audit_security | Scans files for exposed API credentials, SQL injection, and code hazards | | openguild_read_memory | Retrieves architectural rules, user preferences, and past bugfixes | | openguild_write_memory | Saves new engineering standards or decisions directly from the chat session | | openguild_learn | Self-reflects on errors, records lessons, and automatically scrubs secrets | | openguild_search_memory | Searches memory archives using keywords (e.g., "jwt", "timeout", "rate-limit") | | openguild_consult_council | Queries the 16-Mind Council in debate, audit, or consensus mode | | openguild_get_project_context | Inspects detected tech stack, test commands, and invariant rules |

⚙️ Complete CLI Cheat Sheet

| Command / Flag | Purpose | Example | |:---|:---|:---| | npx antigravity-guild | Initialize or update OpenGuild in current project | npx antigravity-guild | | npx antigravity-guild --preset <name> | Use a targeted preset (full, agi, backend, web, mobile, ai-ml) | npx antigravity-guild --preset backend | | npx antigravity-guild --setup-mcp [editor] | 1-Click configure MCP for cursor, claude, antigravity, or all | npx antigravity-guild --setup-mcp all | | npx antigravity-guild --tokens | Benchmark context token consumption and cost savings | npx antigravity-guild --tokens | | npx antigravity-guild --verify [--fix] | Run unified tests, lint, typecheck, and memory check | npx antigravity-guild --verify --fix | | npx antigravity-guild --audit | Scan project for secret leaks and code vulnerabilities | npx antigravity-guild --audit | | npx antigravity-guild --analyze "<idea>" | Auto-generate comprehensive architecture blueprint | npx antigravity-guild --analyze "Real-time sync engine" | | npx antigravity-guild --learn "<insight>" | Record an engineering lesson with secret scrubbing | npx antigravity-guild --learn "Use redis locks for queue jobs" | | npx antigravity-guild --team | Initialize Git-shared team memory in .openguild/ | npx antigravity-guild --team | | npx antigravity-guild --setup-ci | Generate GitHub Actions Council Review workflow | npx antigravity-guild --setup-ci | | npx antigravity-guild --lint-memory [--fix] | Clean duplicate rules and format memory vault | npx antigravity-guild --lint-memory --fix | | npx antigravity-guild --status | Inspect active memory files, council status, and stats | npx antigravity-guild --status | | npx antigravity-guild --export-memory | Backup memory vault to JSON file | npx antigravity-guild --export-memory backup.json | | npx antigravity-guild --import-memory <file>| Restore or merge memory vault from JSON backup | npx antigravity-guild --import-memory backup.json | | npx antigravity-guild --install-hooks | Install Git pre-commit invariant verification hook | npx antigravity-guild --install-hooks | | npx antigravity-guild --dry-run | Preview actions without modifying disk | npx antigravity-guild --dry-run |

🔒 Privacy, Security & Zero-Dependency Guarantee

• 100% Pure Node.js Built-ins: Requires zero npm dependencies. No bloated node_modules trees, no supply chain vulnerabilities.
• Runs Exclusively on Your Machine: Never sends your code, prompts, memory, or metadata to external servers or third-party cloud APIs.
• Automated Secret Scrubbing: Automatically detects and strips API keys, OAuth tokens, and private keys before anything touches memory.
• Safe, Non-Destructive Merging: Never blindly overwrites existing .cursorrules or configuration files; intelligently merges rules while preserving your custom instructions.

❓ Frequently Asked Questions (FAQ)

Does OpenGuild slow down my editor or chat?

No, it speeds it up. Traditional prompt templates dump thousands of lines into the base context, causing slow responses and attention drift. OpenGuild uses an ultra-lean base contract (~770 tokens) and lets the AI query deeper memory via lightweight MCP tools only when needed.

Where is my global memory stored?

Global memory lives in ~/.openguild/memory/ on your machine (C:\Users\<User>\.openguild\memory on Windows). It is organized into clean markdown files (institutional_memory.md, security_standards.md, user_profile.md) that you can inspect and edit anytime.

How do I share memory with my teammates?

Run npx antigravity-guild --team. This creates a .openguild/ folder inside your repository. Commit this folder to Git. Any teammate who pulls the repository will automatically share project architecture decisions and team standards.

Can I use OpenGuild with existing projects?

Yes. OpenGuild was designed specifically for existing, mature codebases. It automatically detects your package manager, test scripts, and directory layout without breaking existing workflows.

📄 License

MIT © Hardik Patel