
Cloud platform providing fast, persistent Linux compute and deployment infrastructure for AI agents.
Dedalus Labs provides a cloud platform for AI agents centered on Dedalus Machines: persistent Linux virtual machines that start in under 250 milliseconds, preserve filesystem and memory across sessions, and bill only for active compute. Developers interact with machines through a CLI and can install system packages, run Docker, access GPUs, and use root access inside a hardware-isolated environment.
The Dedalus SDK lets developers build agentic applications that connect language models to tools via the Model Context Protocol, with support for local and hosted MCP servers, vendor-agnostic model handoffs, and streaming across providers. The platform targets developers building long-running, stateful autonomous agents without managing container infrastructure.
Dedalus Labs differentiates itself from ephemeral sandboxes by offering persistent Linux virtual machines that boot in under 250 milliseconds and preserve state across sessions. This eliminates cold starts and repeated dependency installation while allowing developers to pay only for active compute time rather than idle resources.
The platform combines VM-level hardware isolation with full root access, supporting Docker, GPU workloads, and arbitrary system packages. Its SDK additionally abstracts the complexity of wiring language models to MCP servers, enabling vendor-agnostic model handoffs and streaming tool calls in a small amount of code.
Dedalus Labs charges only for active compute, with idle machines costing zero dollars per hour. Compute is billed per second based on vCPU, memory, and storage consumed while a machine is running.
The company offers a free Hobby tier with a one-time $20 sign-up credit, a Pro tier at $20 per month that includes $20 in compute credits, and an Enterprise tier with custom fleet-level pricing, SSO, RBAC, audit logs, and dedicated infrastructure options.