# MUXI - The Agent Server

> Open-source infrastructure for running AI agents in production. Not a framework. Not a wrapper. A Server.

## What is MUXI?

MUXI (Multiplexed eXtensible Intelligence, pronounced /muk-siː/) is production infrastructure purpose-built for AI agents. Unlike frameworks that you import as libraries, MUXI is a complete server that runs, orchestrates, and manages multi-agent systems at scale.

**Core Philosophy:** Agents are native primitives – declared in YAML, orchestrated at the infrastructure layer, scaled like containers. No frameworks to fight. No queues to wrangle. Just infrastructure that understands what agents do.

## Key Differentiators

- **Infrastructure, not a framework**: Deploy with `muxi deploy` – agents go live instantly
- **Formation-based deployment**: Complete AI systems packaged as deployable YAML units
- **LLM-agnostic**: Use any provider (OpenAI, Anthropic, Google, AWS, Azure, local models)
- **Multi-tenant by design**: Built-in user isolation with per-user credentials and RBAC
- **Self-hosted**: Complete data sovereignty, runs anywhere (on-prem, cloud, air-gapped)
- **Open source**: Elastic License 2.0 (ELv2) and Apache 2.0 – free for commercial use

## Core Concepts

### Formations

A "formation" is a complete AI system packaged as a single deployable unit. Like a Dockerfile for containers, but for multi-agent intelligence. Formations include:

- Agent definitions and configurations
- Domain knowledge (files, URLs, documents)
- MCP (Model Context Protocol) tool connections
- Standard Operating Procedures (SOPs)
- Triggers (schedules, webhooks)
- Settings and environment variables

All defined in declarative YAML. Version-controlled, portable, and deployable with one command.

### The MUXI Stack

Four architectural layers:

1. **SDKs** - Python, JavaScript, Go, Swift, Java clients for programmatic interaction
2. **APIs** - REST endpoints for deployments, chat, streaming, observability
3. **Runtime** - Core orchestration: agents, memory, tools, task execution
4. **Server** - Self-hosted infrastructure for production deployments

### Agent-Native Infrastructure

Agents aren't bolted on – they're native primitives. State persistence, task orchestration, and resource management happen at the infrastructure layer. Agents are declared, deployed, and scaled like containers.

## Key Features

### Deployment & Operations

- **One-command deployment**: `muxi deploy` and you're live
- **GitOps workflow**: Version-controlled YAML, CI/CD integration, instant rollbacks
- **Hot updates**: Zero downtime deployments with automatic restarts
- **Single binary**: No runtime dependencies, runs anywhere

### Intelligence & Execution

- **Intelligent task decomposition**: Complex requests automatically broken into executable subtasks
- **Adaptive workflow creation**: Non-deterministic workflows generated based on context
- **Agent collaboration**: Multi-agent coordination via A2A protocol
- **Multimodal support**: Images, PDFs, audio, video processing out-of-the-box

### Memory & Context

- **Three-tier memory**: Buffer (immediate), persistent (long-term), vector (semantic)
- **User synopsis caching**: LLM-synthesized profiles reduce costs by 80%+
- **Intelligent context management**: Automatic FIFO cleanup with relevance scoring
- **Domain knowledge system**: Pre-load agents with files, docs, policies via RAG

### Integration & Tools

- **MCP Protocol native**: Access 1,000+ tools (GitHub, Slack, Postgres, Stripe, etc.)
- **Multi-user MCP access**: Per-user OAuth tokens and API keys with encryption
- **LLM provider agnostic**: 21 providers, 300+ models via OneLLM
- **Automatic failover**: Provider switching when services fail

### Production Features

- **Multi-tenancy**: Complete session isolation with per-user credentials
- **Real-time streaming**: WebSocket and SSE support for live updates
- **Built-in observability**: 157 event types, stream to 10+ systems
- **Enterprise resilience**: Circuit breakers, exponential backoff, graceful degradation
- **92% cache hit rate**: Semantic caching reduces LLM costs dramatically

### Triggers & Automation

- **Natural language scheduling**: "Check email every hour" or "remind me tomorrow at 2pm"
- **Webhook triggers**: External systems activate agents via HTTP endpoints
- **Event-driven architecture**: Unified orchestration for schedules, webhooks, user interactions
- **Dynamic async execution**: Smart sync/async switching with complexity detection

### Developer Experience

- **Standard Operating Procedures**: Define organizational knowledge agents reason over
- **Artifact generation**: Create documents, spreadsheets, presentations, code files
- **Type-safe SDKs**: Full IDE autocomplete and type checking
- **88.9% test coverage**: Production-ready from day one

## Architecture

MUXI uses a formation-centric architecture where complete AI systems are packaged as YAML definitions. The server manages:

- Formation lifecycle (deploy, update, rollback)
- Multi-tenant isolation and RBAC
- Agent orchestration and task execution
- Memory management across three tiers
- Tool access via MCP protocol
- Event streaming and observability

## Use Cases

- **Customer support systems**: Multi-agent teams handling tickets, refunds, inquiries
- **Internal tooling**: Automating workflows across GitHub, Slack, Jira, etc.
- **Document processing**: Extract, analyze, and generate reports from files
- **Data analysis platforms**: Query databases, generate visualizations, create reports
- **Booking and scheduling**: Coordinate calendars, send notifications, manage reservations
- **SaaS product features**: Embed AI capabilities in your application

## Licensing

- **MUXI Server & Runtime**: Elastic License 2.0 (ELv2)
- **Formations, CLI, SDKs**: Apache 2.0

**You CAN:**
- Use for internal projects and products
- Build and sell products that use MUXI
- Deploy for clients and customers
- Self-host on your infrastructure

**You CANNOT:**
- Offer MUXI itself as a hosted service (SaaS)

## Technical Details

- **Response time**: <100ms average
- **Test coverage**: 88.9% across 11,000+ test lines
- **Observability events**: 157 distinct event types
- **LLM cache hit rate**: ~72% (semantic caching: 92%)
- **Deployment**: Single binary or Docker container
- **Supported platforms**: Linux, macOS, Windows
- **Protocol support**: HTTP/REST, WebSocket, Server-Sent Events

## Comparison with Frameworks

| | MUXI | LangChain/CrewAI |
|---|---|---|
| **Type** | Server infrastructure | Python library |
| **Deployment** | `muxi deploy` | Write deployment code |
| **Configuration** | Declarative YAML | Imperative code |
| **Multi-tenancy** | Built-in isolation | Build yourself |
| **Observability** | Production monitoring included | Add external tools |
| **Async/Scheduling** | Native support | Add Celery/etc. |

Think: Flask is a framework. Nginx is infrastructure. MUXI is the Nginx for agents.

## Business Model

Open-source with sustainable revenue from support services, not paywalls:

- Core stack is free and self-hostable forever
- GitHub Sponsors for priority support
- Professional services for enterprise deployments (coming soon)
- No feature restrictions or bait-and-switch licensing

## Community & Support

- **GitHub**: https://github.com/muxi-ai
- **Email**: hello@muxi.org
- **Twitter**: @muxi_ai

## Who Built This?

Created by Ran Aroussi (@aroussi), a 30+ year tech veteran whose open-source libraries receive 10M+ monthly downloads. Author of "Production-Grade Agentic AI" (600+ page technical book). No VC funding, no exit timeline – built for long-term sustainability.