Problem

AI agents increasingly need access to a real computer—not just a restricted code interpreter—to install packages, execute code, browse the web, run CI pipelines, and perform ML workloads. Traditional sandboxes are often ephemeral, slow to start, expensive to keep warm, and constrained by limited filesystems, package support, permissions, or session lifetimes. Dedalus frames the operational pain as both developer friction and compute waste: its site contrasts traditional sandbox startup times of roughly 2.5 seconds with much faster machine creation, and illustrates a representative configuration costing 60% less than an ephemeral alternative.

The killer use case is giving autonomous coding and task-oriented agents a persistent execution environment that retains filesystem and memory state between sessions. That lets an agent install and use arbitrary system packages, run longer workflows, and safely execute untrusted code without forcing developers to assemble and operate their own container, sandbox, and orchestration stack.

Product / Service

Dedalus Labs provides a hosted infrastructure layer built around fast, persistent Linux machines for AI agents. The current product is exposed through the Dedalus CLI and API: developers create a machine, execute commands, and SSH into it, while the machine retains its filesystem and memory. The company advertises full Linux environments in under 50 milliseconds, root access inside the VM, support for packages and runtimes such as Python, Node, Rust, Go, Java, Ruby, Docker, and GPU/CUDA workloads, and VM-level isolation for untrusted code.

The delivery model is usage-based infrastructure rather than a packaged end-user application. Machines remain persistent, but customers pay for active compute, which is intended to avoid the cost of maintaining always-warm ephemeral sandboxes. The public site currently directs prospective users to join a waitlist, suggesting that the service is still in an early-access or rollout phase even though the underlying interface is designed to be production-oriented.

Market

Dedalus competes in AI-agent infrastructure, particularly the overlap between code-execution sandboxes, cloud developer environments, and compute substrates for autonomous agents. Its positioning is differentiated by combining very fast startup with persistent full Linux machines, broad system-level flexibility, and stronger VM isolation than lightweight browser or language-runtime sandboxes. Relevant alternatives in the broader category include E2B, Modal, Daytona, Fly.io Sprites, CodeSandbox, Vercel Sandbox, and Blaxel; these products are not identical, but they address adjacent needs around running agent-generated code and workloads.

The company was founded in 2025, joined Y Combinator's Summer 2025 batch, reports a 10-person team, and announced an $11 million seed round in October 2025; third-party funding data reports $11.5 million raised in total. Public evidence shows financing and active product development, but no verified customer count or company-disclosed revenue metric. Given the official waitlist, Dedalus is best characterized as an early commercial infrastructure startup rather than a demonstrably scaled business, and the available evidence does not establish whether it is technically pre-revenue.