MultiModel Dev OS

A zero-runtime-dependency workspace standard for governed AI development.
_{Centralize instructions, skills, registries, workflows, adapters, and local gateway metadata in one auditable repository.}

Documentation | Quickstart | Documentation Map | Gateway Architecture | Changelog

Fast Start

Requirements:

• Node.js 20 or newer
• Windows, macOS, or Linux

npm install -g multimodel-dev-os
multimodel-dev-os init
multimodel-dev-os status
multimodel-dev-os verify

Prefer one-off usage:

npx multimodel-dev-os@latest init
npx multimodel-dev-os@latest verify

MultiModel Dev OS is published as multimodel-dev-os on npm. The package has no third-party runtime dependencies.

Why It Exists

AI-assisted development gets messy when every tool has its own rules, prompts, memory, templates, and workflow assumptions. MultiModel Dev OS gives a repository one public source of truth so teams can keep that operating layer structured, portable, and verifiable.

It is designed for projects that need:

• consistent workspace instructions across AI coding tools and editor integrations
• governed prompt, skill, workflow, and permission metadata
• deterministic validation before changes are trusted
• local-first gateway foundations without activating external providers by default
• a clean public repository surface that stays package-safe

What You Get

Workspace Standardization

• Repository-level operating instructions in AGENTS.md
• Adapter assets for common AI coding tool surfaces
• Templates, examples, onboarding helpers, and consistency checks
• Deterministic build output for the standalone CLI

Skill OS Governance

• Skill, prompt, workflow, permission, guardrail, and agent-cluster registries
• RACE+ prompt templates and public business operator templates
• Read-only Skill OS inspection commands
• Validation-only safety model for declarative governance metadata

Gateway Foundation

• Localhost mock gateway runtime with GET /health, GET /v1/models, and POST /v1/chat/completions
• Deterministic mock non-streaming and SSE streaming responses
• Runtime-readable provider, model, local-model, and routing-preset registries
• Dry-run route planning, fallback-chain planning, and resilience simulation
• Preview-only client configuration plans for local OpenAI-compatible testing
• Local, bounded, redacted, in-memory observability and usage accounting

Safety Model

MultiModel Dev OS is intentionally conservative:

• No runtime dependencies: the published CLI uses Node.js built-ins only.
• Validation before trust: strict verification checks cover package hygiene, schemas, registries, docs, gateway contracts, and repository cleanliness.
• Manual publication: npm publication is maintainer-controlled and guarded by scripts/prepublish-guard.js.
• Approval-gated writes: generated plans and adapters are designed to avoid silent global changes.
• Mock-only execution boundary: the v4.2 gateway executes only bundled mock models.
• No hidden provider activation: external providers are metadata-only until a future explicit implementation enables them.
• Private-by-default observability: traces, metrics, usage, and audit events stay local, bounded, redacted, and in memory.

Current Capability vs Direction

Essential Commands

# Inspect the workspace
multimodel-dev-os status
multimodel-dev-os verify

# Start or onboard a project
multimodel-dev-os init
multimodel-dev-os init --template nextjs-saas
multimodel-dev-os onboard analyze

# Inspect Skill OS metadata
multimodel-dev-os skill-os status
multimodel-dev-os skill-os validate
multimodel-dev-os skill-os list skills
multimodel-dev-os skill-os list prompts

# Synchronize adapter files when explicitly approved
multimodel-dev-os adapter sync all --approved

Onboarding analysis and planning are read-only. Applying generated files requires explicit approval.

Documentation

Package and Release Status

• Latest npm package: [email protected]
• License: MIT
• Runtime dependencies: none
• Required Node.js version: 20 or newer
• Optional GitHub Packages mirror: @rizvee/multimodel-dev-os, controlled by GitHub Packages visibility settings

See docs/release-state.md for the current release matrix and historical notes.

Repository Shape

src/        CLI, Skill OS, gateway contracts, runtime, routing, clients, observability
bin/        Generated standalone CLI entrypoint
.ai/        Product schemas, registries, skills, prompts, workflows, and examples
adapters/   Tool-specific adapter assets generated from the shared project standard
docs/       Public manuals, architecture notes, security docs, and release references
tests/      Unit, integration, fixture, and compatibility coverage
scripts/    Build, verification, package, and release-safety checks

Contributing

Contributions are welcome when they preserve the project boundaries:

• keep runtime dependencies at zero unless a future major design decision changes that posture
• keep gateway provider execution explicit, tested, and documented
• avoid hidden writes, global configuration changes, telemetry, and credential loading
• keep public docs factual, polished, and free of private workflow artifacts
• run npm run build, npm test, and npm run verify before proposing changes

Read CONTRIBUTING.md and SECURITY.md before opening larger changes.

License

MIT. See LICENSE.