Architecture

Five layers. Zero lock-in.

weaveIntel's architecture is designed for composability — swap any layer, extend any boundary, deploy anywhere.

Architecture

The five-layer stack

Every layer is independently deployable, horizontally scalable, and fully observable.

1

API Layer

RESTful and WebSocket endpoints for all platform operations. Authentication, rate limiting, and versioned routes. Every capability is API-first — what you see in the UI, you can automate.

REST WebSocket OpenAPI 3.1
2

Orchestration Layer

The Loom Engine — DAG-based workflow execution with agent runtime, task routing, and human-in-the-loop gates. Manages the lifecycle of every AI workflow from creation to completion.

Loom Engine Agent Runtime Task Router
3

Governance Layer

Policy engine, prompt safety, audit trails, and cost controls. Every AI decision passes through governance before execution. Declarative policies in YAML, enforced at runtime.

Policy Engine Prompt Safety Audit Trail
4

Integration Layer

Universal MCP protocol, tool framework, field registry, and data adapters. Connect any external system through a composable, typed interface. One protocol for all integrations.

MCP Tools Adapters
5

Infrastructure Layer

Multi-tenancy, worker pools, caching, metrics, and auto-scaling. The foundation that makes everything above production-ready. Runs on Docker, Kubernetes, or bare metal.

Multi-Tenant Workers Auto-Scale
API Design

Three API surfaces, one experience

Whether you're calling from a browser, a CLI, or another service — you get the same contract.

🌐

REST API

CRUD operations, workflow management, agent configuration, and governance controls. OpenAPI 3.1 with typed request/response schemas.

WebSocket API

Real-time streaming for agent responses, workflow events, and live metrics. Pub/sub channels with automatic reconnection.

🔧

CLI / SDK

Command-line tools for workflow management, deployment, and debugging. TypeScript and Python SDKs for programmatic access.

Principles

Design principles behind every decision

Composability over Monoliths

Every component is a module. Swap the LLM provider, replace the policy engine, or bring your own auth — without touching other layers.

Governance by Default

Security and compliance aren't optional add-ons. Every AI action passes through policy checks, audit logging, and cost tracking from day one.

API-First

Every feature ships as an API first, UI second. If you can't automate it, it doesn't ship. APIs are versioned, documented, and tested.

Observable by Design

Structured logging, OpenTelemetry traces, and Prometheus metrics on every operation. You can't manage what you can't measure.

Technology

Built on rock-solid foundations

TypeScript
Runtime
PostgreSQL
Database
Redis
Cache & Queue
Docker
Deployment
Read the Architecture Guide →