Skip to content

Agents Overview

Agents are the central concept in Aigentic. An agent is an AI-powered assistant that you create, configure, and manage from the console. Each agent has its own model, personality, tools, and conversation history.

What Is an Agent?

Section titled “What Is an Agent?”

An agent in Aigentic combines three things:

  1. An AI model — The language model that generates responses (from Anthropic, OpenAI, xAI, Z.AI, MiniMax, or NVIDIA NIM).
  2. A system prompt — A text document that defines the agent’s personality, instructions, and behavioral constraints.
  3. A toolset — A configurable set of plugins (web search, code execution, file management, persistent memory, and more) and custom skills.

Together, these produce an AI assistant that behaves according to your specifications and can take actions beyond generating text.

How Agents Work

Section titled “How Agents Work”

Isolated Execution

Section titled “Isolated Execution”

Every agent runs in its own isolated environment on the Aigentic platform. Each tenant gets a dedicated gateway container providing:

  • Security — Agents cannot access other users’ data or configurations.
  • Independence — Agents do not interfere with each other. You can run multiple agents simultaneously without conflicts.
  • Dedicated context — Each agent maintains its own conversation history and state.

The Gateway

Section titled “The Gateway”

The Rust backend includes a per-tenant gateway that routes messages between the console UI, agents, and external channels. When you send a message:

  1. The UI sends the message to the backend over a WebSocket connection.
  2. The gateway routes it to the correct agent.
  3. The agent processes the message, optionally invokes tools, and streams the response back.
  4. The gateway forwards the response to the UI and any connected channels.

Conversation History

Section titled “Conversation History”

Agents maintain conversation history within their chat sessions. The Chat page shows all sessions in a sidebar where you can search, rename, clear, export, and delete conversations. History persists across browser sessions. For long conversations, the platform automatically compacts older messages into a summary to reduce token costs while preserving context.

Configuration

Section titled “Configuration”

Each agent has the following configurable properties:

PropertyWhere to Configure
NameSet during creation. Shown on the agent card, chat header, and logs.
AvatarSet during creation or changed from the profile dialog.
ModelSet during creation. Changed by clicking the model link (✎) on the agent card.
System PromptSet during creation or edited from the pencil icon on the card / Soul tab in profile dialog.
PluginsPlugins tab in the agent profile dialog.
SkillsSkills tab in the agent profile dialog.
ChannelsConfigured from the Channels page in the sidebar — not from the agent directly.
WalletWallet tab in the agent profile dialog.

Next Steps

Section titled “Next Steps”