MultiModel Dev OS

Appearance

MultiModel Dev OSOne config. Every AI tool. Zero lock-in.

Stop copy-pasting rules between Cursor, Claude, Gemini, Codex, and VS Code. Standardize your AI pair-programmers with a single portable workspace.

Get Started in 30 Seconds
See It in Action
vs Alternatives
View on GitHub
MultiModel Dev OS Logo
📁

Instant Repo Onboarding

Analyze existing projects, get template recommendations, and bootstrap AI Dev OS configs without breaking anything.

🔄

Universal Adapter Sync

Write rules once in AGENTS.md — auto-syncs to .cursorrules, CLAUDE.md, .vscode/, .gemini/, and 10+ more tools.

🧠

Self-Improving Intelligence

Hash-compressed memory, developer feedback loops, and structured improvement proposals with human-in-the-loop safety gates.

⚡

Caveman Mode (−79% Tokens)

Slash model context footprint by ~79% with compressed shorthand declarations. Save massively on API bills.

🛡️

214+ Quality Gates

Built-in validate, doctor, and verify commands enforce workspace health with strict structural assertions.

🔧

Zero Dependencies

Pure Node.js CLI — no runtime, no build step, no package manager lock-in. Works on Windows, macOS, and Linux.

Install in 30 Seconds ​

bash
npx multimodel-dev-os@latest init

Already have a project? Onboard it safely:

bash
npx multimodel-dev-os@latest onboard analyze
npx multimodel-dev-os@latest onboard recommend
npx multimodel-dev-os@latest onboard apply --approved

Why Not Just .cursorrules or CLAUDE.md? ​

You use Cursor for autocomplete, Claude Code for terminal ops, Gemini for audits. Every tool switch:

  1. Drops context — the next tool doesn't know what the previous one decided
  2. Drifts rules — you update .cursorrules but forget CLAUDE.md, causing confusing failures
  3. Wastes tokens — every prompt resends your full ruleset, and API bills spike

MultiModel Dev OS gives you a single source of truth (AGENTS.md) that auto-syncs to every tool. Change once, every tool updates.

Works With ​

🤖 Codex
🪐 Antigravity
🎯 Cursor
⚡ Claude Code
🧠 Gemini
💻 VS Code
🔌 Cline / Roo
🔌 Aider

See It in Action ​

📁 Repo Onboarding
Analyze your project, get recommendations, and bootstrap configs safely.
~2 min
🔄 Adapter Sync
Mirror rules across Cursor, Claude, VS Code, and Gemini automatically.
~1 min
🧠 Improvement Loop
Feedback → proposals → validate → apply with audit trails.
~2 min
🤝 Agent Handoff
Compile session context and hand off between models.
~1 min
🚀 Release Check
Run 214+ verification assertions and package hygiene checks.
~1 min
⚔️ vs Alternatives
See how MMDO compares to manual rules, prompt packs, and tool-specific configs.

🆕 What's New in v3.5 ​

  • 🛡️ Trusted Registry Signatures — Asymmetric Ed25519 signatures verify remote publisher identity, preventing man-in-the-middle manifest modifications.
  • 🔑 Trusted Key Store — Local keys configuration at .ai/registries/trusted-keys.yaml manages active publisher public keys, statuses, and scope permissions.
  • 📝 Registry Provenance Lockfile — Records synced catalog/manifest SHA-256 hashes, sync times, and verdicts to detect post-sync local tampering.
  • 📊 Structured Verdicts — Generates machine-readable trust verdicts covering registry manifests, catalogs, keys, and signatures for auditability.

Safety & Trust ​

🛡️
214+ Quality Gates
Strict structural verification on every release
👤
Human-in-the-Loop
All writes require explicit --approved flag
🔒
12 Safety Gates
Proposal validation with path boundary checks
📝
Full Audit Trail
SHA-256 hashed apply log for every change

Explore ​

🚀 Quickstart Guide
Deploy adapters and root contracts in under 30 seconds.
⌨️ CLI Reference
Full command reference for init, scan, status, memory, workflow, and more.
📦 Template Gallery
6 production-ready stack configurations for Next.js, WordPress, e-commerce, and more.
🏗️ Architecture
4-layer protocol from root contracts through intelligence engine.
⚡ Cost Optimization
Cut your prompting token bills by up to 79% using Caveman shortcuts.
📦 Distribution Guide
Release checklists, npm publishing, and package hygiene.

For AI Assistants and Coding Agents ​

To support modern AI search, GEO discovery, and developer agents:

  • We provide llms.txt and llms-full.txt at the root of the hosted documentation to allow Large Language Models and AEO utilities to read full workspace specs in single, token-optimized files.
  • The officially frozen protocol guarantees that agents can read and write workspace rules without manual supervision or instruction drift.

Contributing & Community ​

⭐ Star us on GitHub — it helps developers discover this project

Released under the MIT License.

Copyright © 2026-present MultiModel Dev OS team.