Dedalus Labs

About Dedalus Labs

Dedalus Labs provides persistent Linux machines for AI agents, built around the observation that agents do not fit the serverless model: a function that cold-starts, runs, and dies loses everything an agent learned mid-task, while a permanently-on VM bills for hours the agent spends waiting. Dedalus gives each agent a full Linux machine that starts in under 50ms with a persistent runtime that never sleeps, and bills only for active compute — an idle machine costs $0 per hour. The published rates are per-second granularity at $0.04536 per vCPU-hour and $0.01458 per GiB-hour of memory, with storage included in every plan and overage at $0.0001 per GiB-hour, which makes it one of the few agent-compute vendors publishing unit economics openly rather than routing everything to sales. Installation is a single curl-piped script, after which `dedalus start` creates a machine in roughly 12ms and leaves an agent listening. The company frames the product as the substrate underneath agent-to-agent work rather than a sandbox for one-off code execution, and it also runs a store, an ambassador program, and a documented API. Dedalus Labs is Y Combinator-backed and based in San Francisco; the product is currently gated behind a waitlist.

Key Features

Dedalus Labs Pros & Cons

Tags