Your First Agent
A Agent is a persistent AI agent with its own identity, memory, and tools. Unlike disposable chat sessions, an Agent remembers every conversation and builds knowledge over time.
Creating an Agent
Section titled “Creating an Agent”- Open Hivekeep in your browser (default:
http://localhost:3000) - Complete onboarding with Queenie (set up your admin account and first AI provider)
- Click New Agent in the sidebar
- Give it a name, description, and optionally a system prompt
- Choose an AI model from your configured providers
- Start chatting
What makes an Agent?
Section titled “What makes an Agent?”| Property | Description |
|---|---|
| Name | Display name (e.g. “Research Assistant”) |
| Description | What this Agent does |
| System prompt | Instructions, personality, expertise domain |
| Model | Which AI model to use (can be changed anytime) |
| Avatar | Visual identity (auto-generated or custom) |
Key concepts
Section titled “Key concepts”Persistent memory
Section titled “Persistent memory”Every conversation is automatically stored. Agents extract important facts into long-term memory and can recall them later using vector similarity + full-text search.
Session compacting
Section titled “Session compacting”When a conversation gets long, Hivekeep automatically summarizes older messages to stay within token limits. Original messages are always preserved: compacting is non-destructive and reversible.
Agents come with 100+ built-in tools out of the box: web search, memory management, file handling, sub-agent delegation, cron jobs, and more. See Tools for the full list.
Collaboration
Section titled “Collaboration”Agents can talk to each other, delegate tasks to sub-agents, and work on cron schedules. They’re not isolated chatbots, they’re a team.
Next steps
Section titled “Next steps”- Configure environment variables and providers
- Learn about System Prompts for shaping Agent behavior
- Explore Mini-Apps: interactive UIs built by Agents
- Set up Channels (Telegram, Discord, etc.)