An agent in Qorven is a specialist with a job. You give it a name, tell it what to do in plain English, choose which tools it can use, and Qorven does the rest. Your agents run when you need them — in the background, on a schedule, or the moment a message arrives.

Your built-in team

Every Qorven install comes with a ready-to-use team of five agents. You can use them as-is, customise them, or add your own.
AgentRoleWhat it does
PrimeLead AgentYour AI Chief of Staff. Understands your request, decides who handles it, and coordinates the work.
ResearcherIntelligenceDeep web research, source verification, structured briefings.
DeveloperEngineeringReads and writes code, runs tests, opens pull requests, fixes bugs.
WriterCommunicationsDrafts emails, reports, social posts, and anything else that needs words.
Email AgentInbox OperationsMonitors your inbox, triages messages, drafts replies, sends with your approval.

Prime: your lead agent

Prime is the agent you talk to by default. When you send a message, Prime decides what needs to happen and who should do it. Ask Prime anything. If the task needs research, Prime delegates to the Researcher. If it needs code written, the Developer picks it up. If it needs a well-written summary, the Writer handles it. You don’t have to think about which agent to use — Prime does that coordination for you.

What each agent owns

Every agent has its own identity, memory, tools, and model. You can configure all of these:
Name & role descriptionHow Prime identifies and routes work to this agent
InstructionsPlain-English directions — the agent’s “job description”
ModelWhich AI model this agent uses (can differ from other agents)
ToolsWhich built-in capabilities this agent can use
MemoryWhat this agent remembers from past conversations

How agents run

Agents in Qorven don’t run constantly — they activate when there’s work to do. When a message arrives:
  1. The agent loads its instructions, recent history, and relevant memories
  2. It calls the AI model with your message
  3. If it needs to use a tool (search the web, read a file, send an email), it does that and continues
  4. It sends back its reply and saves anything worth remembering for next time
When the conversation ends, the agent goes quiet. No idle processes, no background compute.

Creating your own agents

You can create agents for anything specific to your work — a customer support agent, an agent that monitors your Slack, one that tracks your shipments, or one trained on your company’s internal docs.

Create an agent

Step-by-step guide to building a new agent.

Write good instructions

How to write agent instructions that actually work.

Choose tools

What tools are available and how to control what each agent can do.

Agent memory

How agents remember things — and how to browse and edit those memories.