persistent VM 1

Dedalus Machines are the fastest persistent sandboxes for AI agents. Spin up long-running, stateful agents that never sleep in <50ms. Dedalus Labs provides cloud infrastructure purpose-built for AI agents that need to run for hours, days, or continuously in production. Unlike ephemeral sandboxes that destroy state after each session, Dedalus Machines give every agent its own persistent Linux microVM. Files, installed packages, environment variables, caches, and running processes all survive across sleep and wake cycles. Sleep is free — you only pay for compute when the agent is actually running. Key capabilities: - 50ms cold start with scale-to-zero (no idle charges) - VM-level hardware isolation for running untrusted LLM-generated code - Persistent POSIX-compliant filesystem across sessions - Per-second billing starting at $0.0504/vCPU-hr - GPU-capable VMs for local model inference - Built-in MCP (Model Context Protocol) support with a hosted marketplace of production-ready MCP servers - Multi-model SDK supporting Claude, GPT, Gemini, DeepSeek, Mistral, and 100+ models - Dedalus Auth (DAuth): zero-trust, multi-tenant authentication for MCP servers Dedalus is designed for developers building autonomous coding agents, research agents, CRM agents, monitoring agents, and any stateful AI workflow. Free tier includes $20 sign-up credit and 50 compute hours per month with no credit card required. Backed by Y Combinator (S25) with $11M in seed funding from Kindred Ventures and Saga Ventures.