⚡
SWARIX
Multi-agent AI orchestration framework. Coordinates specialised AI agents across parallel workflows, managing persistent memory, intelligent routing, and cross-agent task delegation.
Active
Python
Multi-Agent
MCP
Overview
SWARIX is a production-grade orchestration layer for coordinating multiple AI agents across complex, multi-step workflows. Rather than relying on a single monolithic AI call, SWARIX decomposes tasks into specialised subtasks and routes them to purpose-built agents running in parallel.
The framework provides a persistent memory substrate, a routing engine for intelligent task dispatch, and an inter-agent communication protocol — enabling agents to share context and coordinate without requiring a central bottleneck.
Framework Architecture
Orchestrator
Central coordination layer that decomposes incoming tasks, determines which agents to invoke, and manages the execution graph across parallel branches.
Memory
Persistent, searchable memory store shared across all agents. Supports semantic retrieval so agents can surface relevant prior context without explicit wiring.
Router
Intelligent routing engine that maps task types to the most appropriate specialised agent based on capability profiles and current load.
Swarm
Parallel agent execution pool. Agents run concurrently and can spawn sub-agents dynamically for deeply nested task hierarchies.
Key Features
- True parallel agent execution — independent subtasks run simultaneously across the swarm
- Shared persistent memory — agents accumulate and retrieve cross-session context
- Dynamic agent spawning — complex tasks can spawn specialised sub-agents at runtime
- Routing hooks allow custom dispatch logic per task type or priority level
- Full execution trace per workflow — agent calls, memory reads, routing decisions, outputs
- MCP-compatible tooling surface for native Claude Code integration
Tech Stack
Python
asyncio
LLM API
MCP Protocol
Vector Store
SQLite