ultra efficient sandboxes for AI agents - compute-disk decoupled AI agent sandbox, 183x microVM reuse/sec, ~2.3 ms cold, ~0.96 ms warm. microVM rotation ~6ms, cuts agent execution costs by 50x–100x..

Disposable, isolated microVMs with container ergonomics, for people and agents: Firecracker VMs from any OCI image, snapshots and fork, jailer and network isolation, web console, MCP server

Fast cold-boot MicroVM sandboxes for AI agents on cocoon

A Linux machine with exact undo and forks for coding agents, as an MCP server: fork the live server, try the fix on the copy, apply it once it works. Firecracker microVMs.

Sandboxed TypeScript runtime for AI coding agents to query AWS — full SDK access with in-sandbox aggregation, credentials flow via IMDS and never enter the VM. Ships as a Claude Code plugin (MCP) and a standalone CLI, backed by Shuru microVMs or Docker.

Easy, fast, isolated microVM sandboxes for AI agents and untrusted code on AWS Lambda MicroVMs — Python SDK, asb CLI, and MCP server.

MCP server for Krova Cloud — let Claude, Cursor, and any MCP client provision and manage Cubes (Firecracker microVMs).

Your agent wants to touch prod. Shadow Run lets it break a clone first: every risky plan runs on a full clone of prod in a microVM on Vultr, and a person approves what it really did, not what it says, before it reaches the real database.

Secure runtime for AI agents. Run AI agents and their tools in Firecracker microVMs with network policies, PII scanning, prompt injection defense, and audit logging. 7 tools: create_sandbox, run_command, read_file, write_file, list_files, kill_sandbox, list_sandboxes. Security presets (none / standard / strict) let you control guardrails per sandbox: restrict outbound traffic to specific domains, redact PII from command output, block prompt injections, and get full audit trails.

Run an AI agent against 5 adversarial attacks (credential leak, metric fabrication, false halt, runaway spend, log falsification), scored on what the harness observed it do '—' never its self-report. Free local self-diagnostic: runs on your own machine, in a temporary working directory, not a security sandbox. Optional paid tier runs the same five server-side in fresh single-use E2B microVMs (isolation enforced by E2B) and returns an ed25519-signed badge. Each attack is PASS, FAIL, or INDETERMINATE; an attack the agent did not demonstrably encounter is never a PASS, and a run containing an INDETERMINATE is never signed. A badge attests one artifact's observed behavior on one date. Not a safety certification.

Self-hosted Docker sandboxes for teams and AI agents: own Docker daemon per sandbox, a URL for every app, sleeps when idle. API + MCP server.

Monorepo for the Krova Cloud JS/TS packages — SDK, CLI, MCP, webhook verifier, n8n node.

Run code in isolated cloud containers. ConTree gives AI agents secure sandboxed execution environments with full root access, network, and persistent images.

Shardflux MCP server (stdio): serverless VMs for your agent. Open a persistent workspace by key, run commands, edit files, use git, a browser and a desktop in it. Uses your scoped Shardflux API key.