Skip to content
Elitea AI Docs

Key concepts

Elitea is built from a small set of concepts that combine to form larger workflows. Each one has its own menu; this page gives you the mental model before you dive in.

flowchart LR
    Credential --> Toolkit
    Toolkit --> Agent
    Skill --> Agent
    Agent --> Pipeline
    Agent --> Conversation
    Pipeline --> Conversation
    Toolkit --> Conversation
    Artifact -.-> Conversation
    Artifact -.-> Agent

Agent

Section titled “Agent”

An agent is a virtual assistant you configure with instructions, an AI model, and optional toolkits. It interprets a task, decides which tools to call, and acts — from searching the web to creating a Jira ticket — without you scripting every step.

A useful agent combines: clear instructions describing its job, one or more toolkits for the services it needs to reach, and a model suited to the task’s complexity. See Agents.

Skill

Section titled “Skill”

A skill is a reusable, Markdown-based instruction set you attach to an agent. Where an agent is a whole assistant, a skill is a focused capability — a checklist, a persona, a domain procedure — that multiple agents can share. Update the skill once and every agent that has it attached picks up the change; invoke it mid-conversation with ~skill-name. See Skills.

Pipeline

Section titled “Pipeline”

A pipeline is a visual, multi-step workflow that connects agents, toolkits, and control-flow nodes (decisions, routers, state updates) into a repeatable process. Use a pipeline when a task needs more than one stage — for example, reading a requirement, drafting a user story, then generating a test case — with each stage’s output feeding the next. See Pipelines.

Credential

Section titled “Credential”

A credential securely stores the authentication details — API keys, tokens, or other secrets — that a toolkit or AI provider needs to connect to an external service. Credentials are created once and can be reused across multiple toolkits and agents, and can be scoped to a project or kept private to you. See Credentials.

Toolkit

Section titled “Toolkit”

A toolkit is a configured integration with an external service (Jira, GitHub, Confluence, and dozens more) or an Elitea-native capability (Artifact storage, a custom OpenAPI connection). Once configured with a credential and a set of enabled tools, a toolkit can be attached to any agent, pipeline, or conversation in the project. See Toolkits and the toolkit catalogue.

Conversation

Section titled “Conversation”

A conversation is an interactive chat session where you talk to models, agents, pipelines, toolkits, MCP servers, and — in team projects — human teammates, all as participants in the same thread. Conversations keep their own history and can be organized into folders, shared, or replayed. See Chat.

Artifact

Section titled “Artifact”

An artifact is a file stored in a bucket that agents, pipelines, and conversations can read from and write to as shared, temporary storage. Buckets have a retention policy and are organized per project, so generated files (reports, exported data, chat attachments) have somewhere durable to live between turns. See Artifacts.