Agent Configuration

Everything you need to know about setting up your agents and making them work together. No technical background required.

What Determines How an Agent Behaves?

Three things shape your agent's behavior. Think of them as layers that build on each other:

Templates — These are your agent's personality and instructions. They tell the agent who it is, how to communicate, and what rules to follow. You write these in the dashboard.
Skills — These are the tools your agent can use. Each skill gives it a new ability, like sending emails, browsing the web, or reviewing code. Install them from the Skills Marketplace.
Secrets — These are API keys and credentials your agent needs to connect to external services. For example, a GitHub token to review pull requests, or a Gmail token to send emails. Add them in the Secrets page.

Each Agent Gets Its Own Space

Every agent runs in its own isolated environment. This means your agents don't interfere with each other — if one agent is busy with a long task, your other agents keep working normally.

Each agent gets dedicated computing resources (200 MHz CPU and 1.5 GB of memory). You don't need to manage this — it's handled automatically.

Multi-Agent Communication

When you have multiple agents, they can talk to each other. This is one of the most powerful features of Autonomis.

Your default agent (usually called "manager", marked with a badge in the dashboard) acts as the orchestrator. When you give it a complex task, it can delegate parts of that task to your specialist agents.

For example, if you ask your manager to "research competitor pricing and write a report," it might send the research part to your researcher agent and the writing part to your writer agent.

Using Group Chat

The easiest way to get agents working together is Group Chat. In a group chat, you can talk to all your agents at once. Use the [INVOLVE:agent-name] pattern to bring a specific agent into the conversation:

[INVOLVE:researcher] — brings your researcher agent into the conversation
[INVOLVE:writer] — brings your writer agent into the conversation

Behind the scenes, agents communicate through a messaging system called Redis Streams. You don't need to set this up — it works automatically when you have multiple agents.

Agent Workspace

Each agent has its own file storage area called a workspace. Think of it as the agent's personal desk where it keeps all its files.

Here's what your agent stores in its workspace:

Conversation history — Records of past chats so the agent can reference previous discussions
Memory files — Notes the agent writes to remember important things long-term (like a personal notebook)
Downloaded or generated files — Any documents, images, code, or reports the agent creates or downloads

Files in the workspace persist across sessions. When your agent restarts, its files are still there.

Vector Memory (Smart Search)

Every agent automatically builds searchable long-term memory. This is one of the features that makes Autonomis agents feel truly intelligent.

Here's how it works:

Automatic indexing — Your agent's memory files are automatically scanned and indexed every hour using AI
Semantic search — Your agent can search its memory by meaning, not just keywords. If you talked about "reducing customer churn" last week, the agent can find it even if you search for "keeping users"
Persistent across restarts — Memory survives restarts and sessions. Your agent won't forget what it learned
90-day retention — Memories are kept for 90 days, then automatically cleaned up to keep things tidy

Chatting with Your Agents

The simplest way to talk to your agents is through the built-in Chat in the dashboard. Just click on any running agent and start typing.

You can also connect your agents to external platforms like Discord, Telegram, Slack, or WhatsApp through Integrations.