Build IntelligentAgent Systems
Production-grade Python framework for building, orchestrating, and deploying multi-agent LLM systems with 100+ ready-to-use capabilities. Open-source, Apache 2.0 Licensed.
pip install alcyoneusProduction-Ready Capabilities
Everything you need to build, deploy, and scale intelligent agent systems — 17 production-hardening phases complete.
StateGraph Engine
LangGraph-inspired cyclic workflow engine with nodes, edges, dynamic routing via Command(goto=...), conditional branches, and subgraphs.
View docs9 Prebuilt Agents & Swarms
ReactAgent, RAGAgent, SwarmAgent, SupervisorTeamAgent, PlanActReflectAgent, StructuredOutputAgent, and AudioAgent ready out of the box.
View docsRealtime Audio & Voice
Gemini Live, OpenAI Realtime, Local Whisper+TTS with audio barge-in interruption, transcript persistence, and SIP telephony integration.
View docsSandboxing & Computer Use
8+ isolated backends (Docker, K8s, Firecracker micro-VMs, Daytona, E2B, Modal) with GUI Computer Use (X11/Wayland/VNC/WebRTC).
View docsOpen Protocols (MCP / A2A / ACP)
Model Context Protocol with Stdio/SSE/WebSocket transports, Starlette-based A2A server with agent cards, and ACP discovery.
View docs50+ Built-in Tools
Search (7 providers + deduplication), browser automation, code interpreter, file I/O, image generation (DALL-E/Imagen/SDXL), and calendar tools.
View docs3-Layer Persistence & Memory
In-memory state, Postgres + Redis checkpointers, QdrantStore, Mem0Store, and dynamic compaction policies for zero context overflow.
View docsEnterprise Policy & Security
9-priority RBAC engine (allow/deny/ask_user), input/output guardrails with tripwire triggers, JWT/mTLS auth, and Vault secrets fallback.
View docsalc CLI & Observability
Unified CLI for scaffolding, HTML/Mermaid graph visualization, replay debugging, OpenTelemetry auto-spans, and 15+ Prometheus metrics.
View docsEngineered for Production Velocity
From deterministic state graphs to realtime voice agents and multi-agent swarms. Explore authentic code patterns from the Alcyoneus OS framework.
import alcyoneus as alc
from alcyoneus.core import StateGraph, Agent, ToolNode, START, END
from alcyoneus.storage.checkpointer import InMemoryCheckpointer
from alcyoneus.prebuilt.tools import safe_calculator, google_web_search
# 1. Define custom domain tools
def check_inventory(sku: str) -> dict:
"""Check live warehouse stock for an SKU."""
return {"sku": sku, "in_stock": True, "warehouse": "us-east-1"}
# 2. Build graph with Agent + ToolNode
graph = StateGraph()
graph.add_node("agent", Agent(model="gemini/gemini-2.5-flash", tool_node="tools"))
graph.add_node("tools", ToolNode([check_inventory, safe_calculator, google_web_search]))
# 3. Add cyclic execution edges
graph.add_edge("agent", "tools")
graph.add_edge("tools", "agent")
graph.set_entry_point("agent")
# 4. Compile with persistence
compiled = graph.compile(checkpointer=InMemoryCheckpointer())
# 5. Run stateful invocation
result = compiled.invoke(
{"messages": [{"role": "user", "content": "Check stock for SKU-9821 and calculate tax at 8.5%"}]},
config={"thread_id": "session-42"}
)The alc CLI Power Tool
Manage your agent lifecycle with a single unified CLI. Scaffold state graphs, render visual diagrams, replay checkpoints, and deploy with confidence.
Real Examples & Capabilities
Explore production-ready examples and subsystems from the Alcyoneus OS repository.
StateGraph + ReAct Agent
Build a complete agent with weather tools, conditional routing, and persistence — the real quickstart example from the Alcyoneus OS repository.
MCP Protocol Integration
Model Context Protocol support with Stdio, SSE, and WebSocket transports, capability negotiation, and 5-minute tool cache.
Realtime Audio Agents
Real-time audio-to-audio agents using Gemini Live, OpenAI Realtime, Azure, and Local Whisper+TTS with barge-in and transcript persistence.
Multi-Agent Swarm
Dynamic multi-agent collaboration with SwarmAgent, shared memory, message passing, hierarchical delegation, and handoff tools.
RAG + Vector Stores
Retrieval-augmented generation with QdrantStore, Mem0Store, memory tools, and dynamic compaction policies for context management.
Sandboxed Execution
Run code safely in Docker, Kubernetes, Firecracker micro-VMs, or local PTY — with GPU passthrough, resource limits, and policy-guarded shell access.
Comprehensive Technical Guides
Over 17 in-depth production guides, architecture specifications, and complete API references.
Quickstart
Build, compile, and run your first stateful AI agent graph in 5 minutes.
Core Patterns
StateGraph, nodes, cyclic edges, conditional routing, and Command(goto=...).
Prebuilt Agents
ReactAgent, RAGAgent, SwarmAgent, SupervisorTeamAgent, and AudioAgent.
50+ Tools Guide
Multi-search (7 providers), browser automation, sandboxed shell, and code interpreter.
Persistence & Memory
Postgres + Redis checkpointing, QdrantStore, Mem0Store, and dynamic compaction.
Streaming & Events
15 streaming event types, SSE heartbeat keep-alives, and low-latency chunking.
Security & Guardrails
9-priority RBAC policy engine, input/output guardrails, JWT/mTLS, and secret vaults.
Protocols (MCP & A2A)
Model Context Protocol (Stdio/SSE/WS), Agent-to-Agent server, and ACP delegation.
Testing Framework
QuickTest one-liners, TestAgent simulation, MockLLM, and isolated test contexts.
Evaluation & CI
LLM-as-judge criteria, trajectory matching, hallucination detection, and simulators.
All Imports Reference
Complete index of all classes, decorators, constants, and utilities in alcyoneus.
Common Gotchas & Fixes
Solutions for recursion limits, checkpoint deserialization, and multi-tenant scoping.
Start Building Today
Install the framework in seconds and start building intelligent agents with 100+ production-grade capabilities.
pip install alcyoneus