E2B

E2B provides isolated cloud machines for AI agents to run code in. Each session gets its own Firecracker microVM with a dedicated kernel rather than a shared container, so an agent can execute shell commands, attach a terminal, and run Python or JavaScript with formatted outputs such as charts, tables and images. Sandboxes can be paused and resumed with a full filesystem and memory snapshot, or forked into up to 100 copies from a saved state. It is built for developers shipping coding agents that edit repositories and run tests, data analysis products with a code interpreter, browser and computer use agents, and research teams that need many parallel environments for benchmarks. The platform exposes sandbox ports over HTTPS, restricts egress by CIDR or hostname, injects secrets at the egress proxy so they never sit inside the sandbox, mounts S3, GCS and R2 storage or persistent volumes, and reports live CPU, memory and disk metrics with logs and signed lifecycle webhooks. SDKs are available for Python and JavaScript. E2B Embed, released under Apache 2.0, lets teams self host on Docker Compose, Terraform on GCP or Kubernetes, and the hosted service runs in US, EU and APAC regions with bring your own cloud options on AWS and GCP. E2B is SOC 2 Type II certified. Pricing is published. Hobby has no monthly fee, includes $100 of one time usage credit, and allows 20 concurrent sandboxes, sessions up to one hour and 10 GiB of storage. Pro is $150 per month plus usage, with 100 concurrent sandboxes, sessions up to 24 hours and 20 GiB of storage; add ons raise concurrency to 600 for $650 per month or 1,100 for $1,150 per month. Enterprise is custom with a $3,000 monthly minimum. Usage is metered at $0.000014 per vCPU per second and $0.0000045 per GiB of RAM per second, with storage free.