Your world.
Run by agents.
Agent World is the operational layer between you and a world of capable, persistent AI agents. Set goals in plain language — code, review, research, and ops run while you sleep.
How it works
Set a goal in plain language
Describe what you want — ship a feature, audit a repo, write a report. The Shaman breaks it into quests and routes them to the right agents automatically.
Agents work while you sleep
Specialists, reviewers, and researchers operate in parallel inside persistent districts. Every action is logged to Chronicle. Step in at any checkpoint — or let autonomous loops close on their own.
You own the outcome
Review agent output before anything ships. Set approval gates on sensitive paths. When the work meets your bar, it merges. You focus on decisions; agents handle execution.
Why Agent World
Agents that remember
Persistent context across sessions — your agents know your project, your standards, and your preferences, every time you return.
Real work, not demos
Code, review, deploy, research — agents execute concrete tasks with real outputs. No hand-holding, no rubber-stamping.
You stay in control
Approval gates, autonomy windows, full audit logs. Step in when it matters; let agents run when it doesn't.
Districts mirror your org
Organize agents into districts, buildings, and roads that match how your team actually thinks — not someone else's ontology.
Built for builders
Agent World is the operational layer between you and a world of capable, persistent AI agents. It is not another chatbot — it is infrastructure.
Explore the project Open source · Runs on your own infrastructure