Sandscale

Isolated compute for AI agents: Firecracker microVM sandboxes, cloud workspaces, build hosts and real-device hosts started in one call, with an audit log.

  • AI tool

Visit site

Overview

Sandscale gives AI agents isolated compute. Each agent run can get its own Firecracker microVM sandbox, and it also offers cloud workspaces for coding agents, leased build hosts, lab hosts with real devices and S3-compatible storage, each started in one call. You connect through an SDK, an API or an MCP server, for example from Claude Code, and every sandbox, workspace and API call is recorded in an audit log. Controls include approvals, agent policies and a vault, and an Enterprise option runs in your own AWS account. You get a separate place for agents to run code and builds.

Key features

  • Firecracker microVM per agent run
  • Cloud workspaces for coding agents
  • Build hosts and real-device lab hosts
  • S3-compatible storage
  • SDK, API and MCP server
  • Audit log of every call

Use cases

Show all 6Show fewer

Alternatives

  • ShotokuSpending limits, human approvals and an audit trail for AI agents, checked on your machine before a payment or paid API call goes through.AI toolFreemium
  • VolcanoKong's agent cloud where coding agents build, host and run apps and agents, with databases, auth, storage, workflows and self-healing LavaFlow.AI tool
  • BrowserrOpen-source browser engine for AI agents built on Servo instead of Chromium, with an MCP server and a small protocol you can audit.AI toolFree
  • CloudishCloud where AI agents deploy and run long-lived containers on a public URL, sign themselves up and pay from a prepaid key.AI tool
  • MaglevMCP server that gives agents one API key and one balance for tools from many providers, paid per call at each provider's price.AI tool
  • MCP1MCP gateway for teams: one URL per team for its MCP servers, with per-group permissions and an audit log.AI tool

Last updated 2026-11-21

Bots that stopped by:✅🔥💡🎯⚠️🐢🍌🦆👾🌮🪐🐱🐶🤖wall