Terminology

Summary of Terminology Relationships

A Worker is the main product, combining behavior (Brain), knowledge (Memory), and tools (Skills) — all of which use Providers via Connectors.
Users consume Workers, Builders create them, and Admins maintain the platform.
EKE and Memory Management power the intelligence and context in all conversations.
Canvas allows workflow design for Specialized Workers, while Universal Worker Builder manages Universal Workers.
Everything relies on structured, governed access through roles and entry points.

Worker

An AI agent designed to perform tasks using LLMs, memory, connectors, and structured logic. Workers can be created, configured, and deployed through the EverWorker platform.

Universal Worker: A general-purpose worker with flexible configuration, suitable for a wide range of tasks. Built using the Universal Worker Builder.
Specialized Worker: A focused, task-specific worker constructed visually in the Canvas interface using nodes and flows.

Provider

An OpenAPI-based definition for external APIs or services. It acts as the technical foundation for integrating external systems into Workers.

Custom Provider: A user-defined OpenAPI Provider not in the central repository.
Central Repository Provider: Curated and managed by EverWorker. Syncable by customers.

Connector

A configuration that connects a Worker to a specific instance of a Provider, managing authentication and access.

Authentication Methods: OAuth (user-level), App Token (system-level), or Hybrid (preferred user token with fallback).
Multiple connectors can exist for a single Provider.

Business User

A non-technical individual who interacts with pre-built Universal Workers via chat. Users don’t modify logic, skills, or memory—they access existing functionality and can upload context (files, URLs) at runtime.

Builder

A technical or advanced user who creates, edits, and configures Workers. Builders use tools like the Universal Worker Builder and Canvas to define skills, behavior, and integrations.

Admin

Responsible for managing infrastructure, Providers, users, and global configurations. Admins govern access control, observability, memory management, and platform operations.

EKE (Enterprise Knowledge Engine)

The intelligence layer of the platform. It enables Workers to integrate and reason over enterprise-specific knowledge, tools, APIs, and memory in a secure and context-aware way.

Memory

Refers to the platform’s system for ingesting, storing, retrieving, and injecting knowledge into LLM conversations.

Vector Memory: Stores content in embedding-based format to support semantic search and relevance.
Memory Items: Individual pieces of content (e.g., PDF, URL, manual text).
Memory Set: A grouping of memory items.
Full Description: Injected into LLMs directly for context.
Embedded Description: Used for semantic matching during retrieval.
Chunking: Automatic splitting of large files for embedding.
Metadata Filtering: Tagging and filtering content by origin, title, creation date, etc.

Session (Universal Worker Chat)

A single conversation thread with a Worker. Sessions have individualized context and memory settings.

Session Context: Includes files, URLs, and toggled skills for that session.
Session Summary: Automatically generated or user-edited summaries for quick recall.
Session History: Log of previous chats, searchable and user-manageable.

Canvas

Visual builder for Specialized Workers. Allows no-code or low-code construction of task flows through a node-based interface.

Node: Building block of a workflow. Types include API Node, Code Executor, Vector Search, Invoke Worker, If/Case, etc.
Builder Chat: Natural language interface for generating nodes.
Raw Tab: Manual code editing for each node.
Status Indicators: Show readiness, execution results, and issues in each node.

Universal Worker Builder

An interface for building Universal Workers using five structured tabs:

Profile: Name, avatar, tags, and public/private toggle.
Knowledge: Memory and knowledge sources.
Brain: LLM configuration, personality, instructions, prompt structure.
Skills: Tools and capabilities integrated into the Worker.
Summary: Final review and deployment screen.

Skills

Capabilities that a Worker can use, composed of:

Specialized Workers: Embedded as callable units within other Workers.
API Providers: Integrated external services (via Connectors).

Skills are selected in the Worker Builder and can be enabled/disabled per session.

Launchpad

The future home screen for Users, which will display active sessions, suggested actions, and personalized guidance for interacting with Workers.

Analytics Dashboard

Tracks performance, ROI, and efficiency of digital Workers.

ROI Metrics: Manual cost/time estimation input by builders used to show hours/money saved.
Utilization Charts: Measure how much each Worker is used.
Failure Rates: Track unsuccessful executions by worker or overall.

API Scraper

A tool that extracts content from websites using XML sitemaps to feed data into memory.

Scraping Jobs: Can be run, paused, or reset.
Sync from Master: Reuse shared scraping configurations.
Performance Settings: Configure concurrency, filters, and timing.

Entry Points

Defined access interfaces based on roles.

User: Worker list and chat only.
Builder: Full creation/edit access to Workers, memory, and Canvas.
Admin: Complete system access.
Dashboard Viewer: Analytics view only.

MCP (Model Context Protocol)

An emerging integration feature to allow dynamic, secure, and consistent communication between Workers and external servers or agents.