Skip to Content
Overview

Welcome to Nexora

Nexora is an open-source, multi-tenant AI agent orchestration platform. Users define agents with personas, tools, and skills. Agents collaborate in real time, execute tools, delegate tasks to sub-agents, and stream results via WebSocket.

Nexora core is MIT licensed and free to use. NexoraCloud  is the self-hosted commercial product with licensing, billing, and enterprise features.

Quick navigation

What Nexora does

FeatureDescription
Multi-tenant authEmail/password, JWT, API keys, TOTP 2FA, invite-only mode
Agent builderVisual React Flow editor — define agents with personas, tools, skills
Sub-agentsAgents delegate tasks to other agents, results aggregated in parallel
LLM routingProvider chain with fallback: Claude, Gemini, OpenAI, Ollama, plus ~25 API providers (OpenRouter, Groq, DeepSeek, Mistral, OpenCode, and more)
Real-time streamingWebSocket + SSE streaming with exponential backoff reconnection
Knowledge base / RAGpgvector-backed file ingestion + semantic search; semantic agent memory; multimodal image input
Tool executionGitHub, GitLab, Telegram, MCP servers, scheduling, SMTP
MarketplaceBrowse, install, and publish skills, tools, personas, and agents from the public registry
Full-text searchSearch across chats and messages

Architecture overview

User → WebSocket → Agent → LLM Provider Tool calls parsed Sub-agents dispatched (Redis-coordinated) Results aggregated → streamed back

The platform is split into six repositories:

  • Nexora — OSS core (this repo, MIT)
  • NexoraCloud — Self-hosted product with licensing and billing
  • NexoraGateway — Internal admin panel + public API gateway
  • NexoraWeb — Marketing site and customer portal
  • NexoraMarketplace — Public package registry for skills, tools, personas, and agents
  • NexoraDocs — This documentation site

Stack

LayerTechnology
BackendPython 3.12, FastAPI, SQLAlchemy async, Alembic
DatabasePostgreSQL 16, Redis 7
FrontendNext.js 15, TypeScript, Tailwind CSS
LLMProvider chain (Anthropic Claude, Google Gemini, OpenAI, Ollama, + ~25 API providers)
Vector / RAGpgvector embeddings (knowledge base + semantic memory)
ContainerDocker Compose + nginx