What Arc Is

Arc is an AI system that can work in two layers.

Layer 1: Your Main Agent

This is the assistant you talk to directly in arc chat, WebChat, or connected channels.

Use it for:

• normal conversation
• ad hoc research or coding help
• planning and execution in a single chat
• creating tasks for background agents when needed

If all you want is a strong AI assistant with tools, this layer may be enough.

Layer 2: Optional Task Agents

These are named background workers such as researcher, writer, or reviewer.

Use them when you want:

• specialized roles
• different models for different jobs
• persistent queued tasks
• multi-step chains like A -> B -> C
• review loops without staying in chat

Why Arc Feels Different

Most tools stop at one conversation.

Arc can also:

• remember global configuration and identity
• queue work in a persistent task board
• move work across multiple agents
• expose capabilities through a dashboard
• grow with local skills and MCP servers

The Simplest Way To Think About It

• The main agent is your front door.
• Task agents are your background workers.
• The gateway runs the dashboard and background processing.
• Skills and MCP are how Arc gets more tools.

When To Stay Simple

Use only the main agent if:

• you mostly chat interactively
• you do not need background queues
• one model and one identity are enough

When To Add Multi-Agent

Add task agents if:

• you want different roles with different prompts
• you want work to continue after you leave chat
• you want reviewer loops or chains
• you want different providers or models per worker