Core concepts

These are the building blocks you will encounter throughout Mojar. Understanding them helps you make the most of the platform.

Agents

An agent is the central object in Mojar. It combines a system prompt, a knowledge base, a choice of AI model, and optional connectors or workflows into a single unit you can chat with, embed on a website, or automate.

Each agent is independent — it has its own documents, settings, and conversation history. You manage all your agents from the Dashboard (accessible via Agents in the sidebar).

See the Quickstart to create your first agent.

Knowledge base and RAG

The knowledge base is the collection of documents you upload to an agent. When you ask the agent a question, it uses Retrieval-Augmented Generation (RAG): it searches your documents for the most relevant passages, then passes those passages to the language model as context so the answer is grounded in your actual content rather than general training data.

Supported formats include PDFs (including scanned documents), Word files, spreadsheets, presentations, plain text, and Markdown. You manage documents from the Knowledge Base section of each agent.

See Uploading documents for details on formats and processing.

Connectors

Connectors let you link an agent to external data sources or services — keeping your knowledge base up to date or triggering actions in other tools.

Connectors are configured per agent under Integrations in the agent sidebar.

See Connectors overview for the full list and setup guides.

Workflows

Workflows let you automate multi-step processes that involve your agent. You can chain together actions — such as retrieving information, calling external APIs, and generating a response — without writing it all as a single prompt.

Workflows are accessed under Workflows in the agent sidebar. See Workflows to learn how to build one.

Embedding

Every agent can be embedded as a chat widget on any website. Mojar generates a short <script> tag (found under Embed in the agent sidebar) that you paste into your page. The widget loads an iframe pointing to your agent; visitors can chat with it directly without leaving your site.

The widget is responsive: it displays as a modal on desktop and as a full-screen experience on mobile.

See Embedding your agent for the embed script reference.

Models

Each agent is powered by a language model from one of four providers: OpenAI, Anthropic, Google, or DeepSeek. The model determines response quality, speed, context-window size, and cost.

You choose or change the model in your agent's Settings.

See Choosing a model for a comparison and guidance on which model fits each use case.