Archivus Platform Architecture

The Six Pillars of Document Intelligence Infrastructure

The Autonomous Agent

Archivus is a fully autonomous AI agent that enterprises can white-label as their own. Connect your tools via MCP, define workflows, and let Archie handle document intelligence autonomously.

Explore the Autonomous Agent →

Platform Overview

Archivus is built on six foundational pillars that together create a comprehensive document intelligence platform. Each pillar represents a core capability that can be leveraged independently or combined for powerful document workflows.

The Six Pillars

1. MCP Protocol Integration

Bidirectional tool connectivity for limitless extensibility

Model Context Protocol (MCP) enables Archivus to both consume and expose AI tools:

  • As MCP Client - Connect to external MCP servers (GitHub, Slack, databases)
  • As MCP Server - Expose Archivus capabilities to other AI systems
  • DAG Integration - Use MCP tools within workflow orchestration

Explore MCP Integration →

2. DAG Orchestration Engine

Production-ready workflow engine for complex document processing

Build sophisticated document workflows with 24 node types:

  • 6 AI Nodes - Summarize, analyze, extract, classify, Q&A, custom prompt
  • 3 Human Nodes - Approval, review, assignment with escalation
  • 1 MCP Node - External tool integration
  • 14 Control Nodes - Conditions, loops, transforms, actions

Explore DAG Orchestration →

3. Proactive Research

Autonomous research with Source of Truth grounding

AI-driven research capabilities that go beyond your document library:

  • Web Research - Search and synthesize external sources
  • Power Mode - Multi-provider parallel search (Pro+)
  • Validation - Fact-check against trusted documents
  • Source of Truth - Ground findings in verified sources

Explore Proactive Research →

4. Rules Engine

Self-learning document organization with hierarchical rules

Automate document organization through intelligent rules that learn from your behavior:

  • 30+ Condition Fields - Match on document type, vendor, amount, AI categories
  • 7 Action Types - Route, tag, assign, notify, trigger workflows
  • AI Learning - Suggests rules based on detected patterns
  • Hierarchical Scoping - Tenant → Organization → Workspace → Project → User

Explore Rules Engine →

5. Multi-Tenant Architecture

Enterprise-grade isolation with zero compromise

5-layer security ensuring complete data separation:

  • Layer 1: Authentication - JWT + API Key verification
  • Layer 2: Authorization - Role-based access control
  • Layer 3: Tenant Context - Request-level isolation
  • Layer 4: Service Layer - Business logic validation
  • Layer 5: Database RLS - PostgreSQL Row-Level Security

Explore Multi-Tenant Architecture →

6. Hybrid Analytics

DuckDB-powered analytics for real-time insights

High-performance analytics layer combining OLTP and OLAP:

  • DuckDB on S3 - Parquet files for analytical queries
  • 10-100x Faster - Aggregations compared to PostgreSQL
  • Zero Impact - Analytics don’t affect operational workloads
  • Cost Effective - Pay only for storage, not always-on compute

Explore Hybrid Analytics →

Architecture Diagram 

┌─────────────────────────────────────────────────────────────────────┐
│                        Archivus Platform                              │
│                                                                       │
│  ┌─────────────────────────────────────────────────────────────────┐ │
│  │                     API Gateway Layer                           │ │
│  │    Authentication │ Rate Limiting │ Tenant Context │ Routing    │ │
│  └───────────────────────────┬─────────────────────────────────────┘ │
│                              │                                       │
│  ┌───────────┬───────────────┼───────────────┬───────────────┐     │
│  │           │               │               │               │     │
│  │  ┌────────┴────────┐ ┌────┴────────┐ ┌────┴────────┐     │     │
│  │  │  MCP Protocol   │ │    DAG      │ │  Research   │     │     │
│  │  │   (Pillar 1)    │ │  Engine     │ │  Engine     │     │     │
│  │  │                 │ │ (Pillar 2)  │ │ (Pillar 3)  │     │     │
│  │  │ ┌─────────────┐ │ │             │ │             │     │     │
│  │  │ │MCP Client   │ │ │ 24 Node     │ │ Web Search  │     │     │
│  │  │ │MCP Server   │ │ │ Types       │ │ Validation  │     │     │
│  │  │ └─────────────┘ │ │ Human-Loop  │ │ Grounding   │     │     │
│  │  └────────┬────────┘ └──────┬──────┘ └──────┬──────┘     │     │
│  │           │                 │                │           │     │
│  │  ┌────────┴─────────────────┴────────────────┴────────┐ │     │
│  │  │           Core Services Layer                       │ │     │
│  │  │   Documents │ Search │ Chat │ Collections │ AI      │ │     │
│  │  └────────────────────────┬────────────────────────────┘ │     │
│  │                           │                               │     │
│  │  ┌────────────────────────┼────────────────────────────┐ │     │
│  │  │    Multi-Tenant Layer (Pillar 4)                     │ │     │
│  │  │   Tenant Isolation │ RLS (40+ tables) │ RBAC         │ │     │
│  │  └────────────────────────┬────────────────────────────┘ │     │
│  │                           │                               │     │
│  │  ┌────────────┬───────────┴───────────┬────────────────┐ │     │
│  │  │            │                       │                │ │     │
│  │  │ PostgreSQL │       Redis          │    DuckDB      │ │     │
│  │  │  (OLTP)    │      (Cache)         │   (OLAP)       │ │     │
│  │  │            │                       │  (Pillar 5)    │ │     │
│  │  └────────────┴───────────────────────┴────────────────┘ │     │
│  │                                                           │     │
│  └───────────────────────────────────────────────────────────┘     │
└─────────────────────────────────────────────────────────────────────┘

Tier Availability

Pillar Free Starter Pro Team Enterprise
MCP Protocol - - - Full Full
DAG Orchestration - - - Full Full + Custom
Proactive Research - - Full Full Full
Rules Engine 5 rules 25 rules 100 rules 500 rules Unlimited
Multi-Tenant Basic Basic Full Full Full
Hybrid Analytics - - - Full Full

Integration Points

API-First Design

All platform capabilities are accessible via REST API:

  • 150+ Endpoints - Complete platform access
  • OpenAPI Spec - Full API documentation
  • Webhooks - Real-time event notifications
  • SDKs - JavaScript, Python, Go

Extensibility

Build on top of Archivus:

  • MCP Servers - Add custom tools
  • Custom DAG Nodes - Extend workflow engine
  • Webhooks - Integrate with external systems
  • Embeddable Components - Add to existing apps

Performance Characteristics

Component Latency Throughput
API Requests 50-200ms 1000+ RPS
Document Search 300-800ms 100+ QPS
AI Operations 2-15s Concurrent
DAG Execution Variable Parallel nodes
Analytics Queries 100ms-5s OLAP scale

Getting Started

For Developers

  1. API Reference - Explore 150+ endpoints
  2. DAG API - Build workflow automations
  3. Code Examples - Integration patterns

For Architects

  1. Multi-Tenant Architecture - Security model
  2. Deployment Options - Infrastructure choices
  3. Enterprise Overview - White-label and BYOB

Ready to build on Archivus? Explore the API →

Table of contents