Platform Architecture

Sharely.ai is built on a modern, scalable architecture designed to deliver intelligent AI agents powered by comprehensive knowledge management.

Platform Overview

Sharely.ai provides a complete infrastructure for building, deploying, and managing AI agents with access to your organization's knowledge. The platform consists of three foundational components that work together seamlessly.

What This Section Covers

The following sections provide an overview of:

  • System Architecture Diagram - Visual representation of the complete platform
  • Knowledge Management - How content is ingested, processed, indexed, and made available with RBAC capabilities
  • Spaces - User session management, context persistence, and conversation state
  • Agent Distribution - Delivery mechanisms (Hosted Agents and Web Control)
  • Integration Points - How components communicate and extend functionality
  • Data Flow - How information moves through the system
  • Security Model - Authentication, authorization, and data protection

Core Components

Sharely.ai is built on three foundational components:

1. Knowledge Management

The Knowledge Management component provides comprehensive content ingestion, organization, and retrieval with enterprise-grade access control.

Content Ingestion & Processing

  • Knowledge Connectors - OneDrive, Google Drive, websites, files, direct text
  • Processing Pipeline - Document parsing, chunking, embedding
  • RAG Foundation - Vector search and semantic retrieval for intelligent content access

Taxonomies & Categories

Taxonomies provide hierarchical organization that enables users to navigate and discover knowledge through structured exploration:

  • Taxonomies - Hierarchical structures that organize categories by role and language, enabling users to browse knowledge in a way that's relevant to them
  • Categories - Parent-child hierarchical organization with metadata that groups related knowledge items
  • Discovery-Oriented Navigation - Users can explore taxonomies to discover adjacently related concepts and find resources they didn't know to search for
  • Multilingual Support - Taxonomies can be organized by language (en, es, pt, etc.) to provide localized knowledge navigation
  • Published Status - Control which taxonomies are active and visible to users

How Taxonomies Enable Discovery: Taxonomies provide a browsable, hierarchical view of knowledge that helps users:

  • Navigate from broad topics to specific resources
  • Discover related concepts adjacent to what they're currently viewing
  • Understand the relationship between different knowledge items
  • Find content they wouldn't have thought to search for

RBAC Capabilities

Role-Based Access Control is an optional capability that filters knowledge experiences based on user roles:

  • Roles - Define user permissions with UUID identifiers and external customerRoleId for system integration
  • Role Assignment - Assign roles to taxonomies, categories, and individual knowledge items for granular access control
  • Filtered Experiences - When RBAC is enabled (workspace.rbacStatus === ACTIVE), Browse, Search, and AI Chat experiences show only knowledge assigned to the user's role

RBAC-Filtered Experiences:

When RBAC is enabled, all knowledge experiences are filtered by the user's role and language:

  1. Browse - Navigate taxonomies and categories filtered to show only content assigned to the user's role and language
  2. Search - Perform keyword and semantic search across only the knowledge assigned to the user's role and language
  3. AI Chat - Engage in conversational AI that accesses only the knowledge assigned to the user's role and language

Learn more: Knowledge API | Taxonomy API | Categories API | Roles API

2. Spaces

Spaces provide personalized AI assistant environments for each user with persistent context and conversation state.

  • User Agents - Personal AI assistants for each user
  • Space Structure - Main thread, goal threads, storage
  • Context Management - Conversation history and state persistence
  • Collaboration - Human-AI handoff and team integration
  • Role-Based Context - Spaces respect RBAC when accessing knowledge, ensuring users only interact with permitted content

Learn more: Spaces API | Spaces Concept

3. Agent Distribution

The Agent Distribution layer provides flexible deployment options for delivering AI capabilities to your users:

Hosted Agents

Fully managed deployment on the Sharely.ai platform. Perfect for:

  • Quick launches without development effort
  • Sharing AI agents via simple links
  • Creating customer-facing AI experiences
  • Testing and prototyping

How it works: Sharely.ai hosts the complete user interface on our domain. You configure the agent, knowledge, and journeys through our platform, then share links to your users.

Web Control

Embeddable JavaScript component for custom applications. Ideal for:

  • Seamless integration into existing web applications
  • Custom branding and user experiences
  • Integration with your authentication system
  • Role-based access control (RBAC)
  • Enterprise and SaaS applications

How it works: You embed a JavaScript component into your application, maintaining full control over hosting, branding, and user experience. Web Control connects to the same knowledge management and spaces infrastructure as hosted agents, but renders within your application.

Why Web Control exists: While hosted agents provide instant deployment, many organizations need AI capabilities integrated directly into their existing applications with their own branding, authentication, and user experience. Web Control bridges this gap by providing the full power of Sharely.ai's knowledge management and AI capabilities as an embeddable component.

Learn more: See the Web Control documentation for complete integration guide, authentication patterns, and API reference.

Bring Your Own Agent (Coming Soon)

Deploy custom AI agents built on any framework while leveraging Sharely.ai's infrastructure. Perfect for:

  • Domain-specific agents with specialized tools and workflows
  • Custom agent loops and reasoning patterns
  • Integration with industry-specific systems and databases
  • Framework flexibility (LangChain, LangGraph, Temporal Workflows, custom implementations)
  • Full control over agent logic and behavior

How it works: Build your agent using any framework or SDK, deploy the open-source Custom Agent Server (MIT License), and connect to Sharely.ai's knowledge management and distribution infrastructure. Your agent maintains complete autonomy while benefiting from Sharely.ai's RAG, RBAC, and deployment capabilities.

Why it exists: Through working with customers across industries, we learned that AI agents need domain-specific tools and agent loops to deliver real value. While our out-of-box agent serves general knowledge access well, specialized domains—legal research, medical diagnosis, code review, financial analysis—require custom capabilities. As foundation models evolved to better support tools and agentic workflows, we're enabling developers to deploy their own agents to any cloud platform while leveraging our knowledge infrastructure.

Learn more: See the Bring Your Own Agent documentation for architecture details, use cases, and early access information.

Knowledge Management & Integration

Knowledge Synchronization

Programmatically sync knowledge from multiple source systems into Sharely.ai using the API. Perfect for:

  • Aggregating siloed knowledge from CMSs, file storage, and databases
  • Automating content updates from WordPress, Contentful, Strapi
  • Maintaining synchronized knowledge across systems of record
  • Building reliable, idempotent sync workflows with Temporal

How it works: Define your desired knowledge state in a YAML configuration file (source of truth), then run a sync script that reconciles Sharely.ai's current state with your configuration—adding missing items, removing orphaned content, and applying role-based access control automatically.

Why it exists: Organizations often have knowledge spread across multiple systems—Azure Blob Storage, WordPress blogs, video libraries, podcast platforms, and external databases. Knowledge Synchronization enables you to unify this siloed content into Sharely.ai's knowledge base, making it accessible to AI agents that can then deliver unified insights to end users.

Learn more: See the Knowledge Synchronization documentation for complete implementation guide, YAML configuration patterns, and CMS integration examples.

External Integrations

  • Knowledge Connectors - OneDrive, Google Drive, web crawlers
  • Your AI Agent - Custom agent integration with tools and MCP protocol
  • Your App Web UX - Embedding Sharely.ai into your application

API & Integration Resources

Explore the complete platform capabilities:

Core APIs

Concepts & Guides

Questions?

If you have specific questions about the platform architecture, please contact us at support@sharely.ai.