E2B is a sandbox platform that gives each AI agent session its own isolated machine. The page describes a microVM per session with a shell, filesystem, browser and network access, along with an API for running code, controlling network traffic, keeping files, holding secrets, watching metrics, and pausing, resuming, snapshotting or forking the sandbox.
The runtime is open source under Apache 2.0, and the company says it is SOC 2 Type II. Agents built with frameworks such as the OpenAI Agents SDK or Claude Code can run inside these sandboxes for coding, data analysis, computer use, evaluations and reinforcement learning rollouts.
E2B is aimed at businesses building AI agents. Teams can use the E2B cloud in US, EU or APAC regions, or deploy inside their own AWS, GCP or Azure account, which the page calls BYOC. Model calls, MCP tools, buckets, repositories and internal APIs can be reached from the sandbox under an egress policy the customer sets. Secrets are passed to a sandbox through an egress proxy so the agent can use a key without reading it.
The page lists a free Hobby tier with a one-time $100 usage credit, sessions up to one hour and 10 concurrent sandboxes. The Pro plan costs $150 a month plus usage, with sessions up to 24 hours and 100 concurrent sandboxes, expandable to 1,100 with add-ons. Enterprise pricing is custom and covers committed usage, BYOC deployment and higher concurrency. Usage is billed per second of a running sandbox, based on vCPU, memory and disk rates listed on the pricing page.
The page cites over 100K teams and more than 2 billion sandboxes started. It also says that 94 of the Fortune 100 have signed up.
