E2B provides AI agents with secure, instant-on machines and real-world tools. From a two-person open-source project to the platform 94 of the Fortune 100 have signed up for.
The open-source, secure environment.
We built the cloud layer that AI agents were missing.
About E2B
The tools built for human developers were never meant for agents. We recognized this early.
When AI models started writing and running their own code, developers reached for the tools they already had: Docker containers, Kubernetes, a VM somewhere. But those were built to run software you define ahead of time, not code an agent generates at runtime.
We saw the gap before most enterprises did. An agent needs its own machine: isolated, fast, and containing real-world tools, just as a person needs a laptop or a terminal. So we built one that spins up in milliseconds.
Our vision is much bigger, though.
What if AWS was purpose-built for agents?
These numbers came from developers at the world's most ambitious companies. They needed secure, fast, reliable compute for their agents, and they chose E2B to run it.
94
Of the Fortune 100 signed up
Ninety-four of the hundred largest US companies have signed up for E2B. Their security reviews shaped the platform.
1B+
Sandboxes started
A billion sandbox starts and counting, across coding agents, data analysis, computer use, and RL training runs.
10M+
Monthly SDK downloads
Python and JavaScript SDKs, npm and PyPI combined. Up from under a million a month in mid-2025.
A self-improving cloud for agents
We started by giving agents a machine. The next step is a cloud that gets better the more it's used.
The sandbox is already where agents do most of their work. It is the runtime that connects any model to real tools: skills, data connectors, and the code itself. It morphs to the workload: more CPU when the agent needs it, gone when it does not, and it runs with any LLM you bring.
Because every agent does its work here, we can see how that work actually gets done. That becomes a flywheel: agents run on E2B, their workflows generate data, the environment learns, and the agents get better, all without your data ever leaving your own cloud.
The agent shouldn’t adapt to the limits of its environment. We’re building an agentic cloud that adapts to the needs of an agent on the fly.
Join usJun 2026
A billion sandboxes. On June 3, we joined the ‘tres commas’ club. Combined SDK downloads climb from 50 to 60 million in 33 days, peaking at nearly 435,000 in a single day.
Aug 2025
E2B package stargazer count climbs to 10K.
Jul 2025
Series A, $21M. Led by Insight Partners, with Decibel, Sunflower, and Kaya, plus angels including Docker's former CEO. 88 of the Fortune 100 are signed up at the time.
2025
Growing fast. The npm SDK crosses its first million installs in March. We sign our first seven-figure enterprise contract, and onboard major customers while supporting the open-source growth. Monthly sandbox volume crosses 15 million.
Oct 2024
Agents get a screen. Desktop Sandbox (beta) launches: a full graphical Linux environment for computer-use agents, shipped right as the first models learn to click.
Oct 2024
v1.0 and a seed round. E2B SDK v1.0 ships. The same week, $11.5M seed led by Decibel.
Mar 2024
40,000 sandboxes a month. Code interpreting becomes the wedge: AI data analysis and visualization in a headless notebook environment.
Feb 2024
The GitHub scare. GitHub accidentally suspends our entire open-source org. For a company built on open source, a genuine mini heart attack. It is restored, and we never take distribution for granted again.
Sep 2023
Pre-seed. ~$2.5M led by Kaya VC, with angels including the founders of Supabase, Deepnote, and Better Stack.
Aug 2023
First SDK ships. The E2B Python SDK lands on PyPI. The current npm SDK line (Feb 2024) and the Code Interpreter SDK (Apr 2024) follow. These are the packages developers still build on today.
Apr 2023
E2B hits 1000 stars.
Mar 2023
Born from a burnout break. Two burned-out founders take a two-week break from DevBook to play with GPT-3.5, and give an agent tools to run code inside their old sandbox. A demo goes viral, Greg Brockman of OpenAI reshares it, and within days they drop the agent and bet the company on the sandbox underneath. The project gets a name: E2B, or English to bits.
2022
The precursor. Before E2B, there was DevBook: in-browser interactive docs and playgrounds for developers. The sandbox tech that powered the product became the foundation of everything that followed.
Before 2022
Two friends and co-founders, Tomas and Vasek, hack away on different projects together, with varying degrees of success. They never make it to YC despite trying multiple times.
A small, cracked team building at the frontier of agentic AI.
Vasek Mlejnsky
CEO & Co-founder
Tomas Valenta
CTO & Co-founder
How about you?
Apply now!
