Agent configuration - Hyperagent Knowledge Base

Every agent in Hyperagent has eight configuration tabs. This page documents every tab and its fields.

Overview

The landing tab when you open agent settings. Displays a summary view of the agent.

Field	Description
Agent name	Display name (read-only here; editable in Identity tab)
Description	Short summary of the agent's purpose (read-only here)
Thread count	Number of threads this agent has participated in
Last active	Timestamp of the most recent activity
Created date	When the agent was first created

Everything that defines who this agent is. The system prompt lives here alongside the name, not as a separate concept.

Field	Description
Name	The agent's display name. Shown in agent picker, sidebar, and thread headers.
Icon	Visual identifier. Choose from built-in icons or emoji.
Description	A short summary of the agent's role and purpose. Shown in the agent picker when selecting an agent for a thread.
System Prompt	The detailed instruction set that shapes the agent's behavior. This is the "job description." Write what the agent should do, how it should think, what format to use, what to avoid. There is no length limit, but focused prompts work better than sprawling ones.

A timeline of the agent's usage history.

Field	Description
Recent threads	List of threads the agent has participated in, sorted by recency
Usage frequency	How often the agent is used over time
Invocation history	Log of how the agent was triggered (thread, Slack, scheduled, etc.)

Controls the underlying AI model and its behavior settings.

Field	Description
Model ID	Which AI model to use (e.g., Claude Opus, Claude Sonnet). Different models have different capabilities, speeds, and costs.
Effort level	How much processing effort the model applies. Higher effort produces more thorough results but takes longer and costs more.
Thinking tokens	Whether the model should use extended reasoning. When enabled, the model "thinks" before responding, improving quality for complex tasks.
Budget	Maximum token spend per response or per run. Prevents runaway costs on complex or recursive tasks.
Subagent model	The model used when the agent delegates work to a subagent. Often set to a faster/cheaper model than the main agent.

All seven trigger methods plus Live Mode configuration.

Invocation Type	Description	Stateful?
Thread	Default. Open a thread and select the agent.	Per thread
Slack	Agent responds to @mentions in connected Slack channels.	Per thread
Telegram	Agent paired with a Telegram bot via BotFather.	Per thread
Scheduled	Cron-based timer. Each run creates a new thread. Configure schedule, timezone, prompt, and delivery channel.	No (new thread)
Webhook / API	HTTP POST endpoint. Each request creates a new thread. Returns an endpoint URL and auth token.	No (new thread)
Email	Dedicated email address. Incoming emails trigger runs. Subject + body become the prompt.	No (new thread)
Live Mode	Interval-based with a checklist. Stays in the same thread. Silent by default; speaks up when something changes.	Yes (same thread)

Coming soon: SMS and MCP Server.

Toggle built-in tools and connect third-party services.

Tools are built-in capabilities toggled on or off:

Category	Examples
Search & Research	Web search, image search
Content Creation	Image generation, video generation, audio generation, avatar videos
Data & Code	Code execution, table creation, file management
Documents	Document creation, webpage publishing, slide creation
Maps & Location	Map generation, geocoding, directions, Street View
Communication	Suggest follow-ups

Integrations are connected third-party services requiring OAuth or API keys:

Integration	Capabilities
GitHub	Read repos, create issues, commit code
Slack	Post messages, read channels
Google Sheets	Read/write spreadsheet data
Gmail	Send/read emails
Linear	Manage tasks and projects
Notion	Read/write pages and databases
And more...	New integrations are added regularly

Attach, pin, and manage skills for this agent.

Concept	Description
Pinned skills	Always loaded into context. Use for skills the agent needs in every conversation.
Available skills	Loaded when the agent determines they are relevant to the current task.
Discoverable skills	Not attached to the agent directly, but available through search if the agent looks for them.

See How Agents Load Skills for the full loading mechanism.

Everything the agent knows beyond its base training.

Field	Description
Memories	Learned facts, preferences, and context. Managed via importance levels (1-5) and categories.
Memory Scope	Controls which memories the agent sees.
Documents	Reference materials attached to the agent. These are documents the agent can access in any conversation.
Context files	Uploaded files that provide background information. PDFs, text files, spreadsheets -- any file the agent might need to reference.
Knowledge Access presets	Controls how knowledge is shared in team workspaces. Determines whether memories and context are shared across team members or kept private.