Learnings from building virtual machines for long-horizon agents. MicroVMs, agent infrastructure, and running AI at the frontier.
Run Harbor and Terminal-Bench evaluations in the cloud without a local Docker daemon: install, run trials in parallel in isolated sandboxes, and read the results.
Run Cursor agents on a machine you configure, with state kept between turns and no compute bill while the agent waits.
Phaser built a Game Agent that turns a one-sentence prompt into a playable game, with design, art, code, and sound. Every game build runs inside a Superserve sandbox.
Keep real API credentials out of agent sandboxes. Superserve Secrets swaps a stand-in token for the real key at the network edge, only for allowed hosts.
Anthropic runs the agent loop, and tools execute in Superserve sandboxes.
A practical Hermes setup walkthrough for choosing a host, wiring up a model and Slack, and keeping the gateway running reliably.
A practical OpenClaw setup walkthrough for installing OpenClaw, wiring it to Slack, and keeping the Gateway running reliably.
A working developer's guide to picking the right OpenAI Agents SDK sandbox execution backend without the marketing.
