moxxy

moxxySelf-hosted Multi-Agent AI Framework

Run local AI agents with sandboxed workspaces, 85 built-in primitives, and multi-channel integrations. Built in Rust.

Get Started
View on GitHub
Moxxy
🤖

Multi-Agent Isolation

Each agent has its own sandboxed workspace, persona, markdown memory journals, encrypted vault grants, and configurable primitive allowlists.

🧠

85 Built-in Primitives

Filesystem, Git, shell, HTTP, browsing, memory, vault, webhooks, heartbeats, MCP, and hive orchestration -- all with fine-grained allowlists.

📱

Telegram & Discord Channels

Connect agents to Telegram and Discord. Pair users, route messages, and let agents respond autonomously.

🔧

Extensible Skill System

Markdown files with YAML frontmatter define reusable skills. Quarantine-first approval model. MCP server integration for external tools.

🔒

OS-Level Sandboxing

sandbox-exec on macOS, bwrap on Linux. Workspace confinement, command allowlists, domain-gated HTTP, and WASI plugins via wasmtime.

⏰

Heartbeat Autonomy

Cron-based heartbeats enable proactive agent behavior. Agents wake up, execute tasks, and go back to sleep on schedule.

🐝

Hive Orchestration

Queen/Worker hierarchy with task boards, signals, voting consensus, and collective decision-making across agent swarms.

📦

Self-Hosted & Data Sovereign

Everything runs locally. Single SQLite WAL database. Your agents, your infrastructure, your rules.

Quick Start ​

Get up and running in minutes:

bash
# Install the CLI via npm
npm install --global @moxxy/cli

# Run interactive setup (provider, API keys, gateway download)
moxxy init

# The gateway starts automatically -- your API is live at http://127.0.0.1:3000

Architecture Overview ​

┌──────────────────────────────────────────────────┐
│              moxxy-gateway (Rust/Axum)           │
│              REST API on port 3000               │
├──────────────┬───────────────────────────────────┤
│  Telegram    │  Discord                          │
│  Channel     │  Channel                          │
├──────────────┴───────────────────────────────────┤
│           ReAct Loop (LLM + Primitives)          │
│       LLM → Primitive Call → Result → LLM        │
├──────────────────────────────────────────────────┤
│  Agent 1          │  Agent 2          │  Agent N  │
│  ┌─────────────┐  │  ┌─────────────┐  │          │
│  │ workspace/  │  │  │ workspace/  │  │   ...    │
│  │ memory/     │  │  │ memory/     │  │          │
│  │ config.yaml │  │  │ config.yaml │  │          │
│  │ allowlist   │  │  │ allowlist   │  │          │
│  └─────────────┘  │  └─────────────┘  │          │
├──────────────────────────────────────────────────┤
│  moxxy.db (SQLite WAL) │ vault.key (AES-GCM)    │
└──────────────────────────────────────────────────┘

What Can Moxxy Do? ​

  • Automate repetitive tasks -- Agents execute shell commands, manage files, and interact with Git repositories
  • Monitor systems -- Heartbeats wake agents on schedule to check APIs, logs, and services
  • Orchestrate agent teams -- Hive mode coordinates Queen and Worker agents for complex multi-step workflows
  • Build conversational bots -- Deploy agents to Telegram and Discord with user pairing and message routing
  • Develop AI applications -- Full REST API with SSE events, bearer token auth, and 60 event types

Documentation Sections ​

SectionDescription
Getting StartedInstallation, setup wizard, and first agent
CLI ReferenceAll CLI commands and gateway management
Core ConceptsAgents, primitives, skills, memory, and hive orchestration
ConfigurationLLM providers, vault, and sandbox settings
ChannelsConnect to Telegram and Discord
API ReferenceREST API with authentication, agents, runs, events, and more
AdvancedOS-level sandboxing and webhook integrations

Community ​

License ​

Moxxy is open source under the MIT License.

Open source · Self-hosted · Data sovereign

© 2026 Moxxy. Built with Rust.