Obelisk

Obelisk houses persistent long-running AI agents called Scribes. Each Scribe runs on its own dedicated VM, giving teams a simple way to deploy, monitor, and manage free-form agents without building their own infrastructure layer. You provision an agent, name it, choose its region and size, and Obelisk handles the lifecycle around it. Unlike rigid workflow tools, Obelisk is built for agents that need to reason, adapt, and act in open-ended environments. A Scribe runs the Brainstorm runtime, which supports tool use across shell, browser, filesystem, and plugins, allowing agents to operate with broad capability instead of being boxed into predefined steps or output formats. Obelisk provides the operational layer around that runtime: a server-rendered UI and API for provisioning and control, real-time telemetry and heartbeat tracking for visibility, and clean teardown when an agent is no longer needed. Everything is scoped to an organization, so agents, API keys, and audit trails stay isolated and manageable. In practice, Obelisk gives companies and non-technical teams a way to run safe, autonomous agents as durable infrastructure.