Docker & E2B partner to introduce MCP support in E2B Sandbox

Today, we’re launching native MCP support in the E2B Sandbox. We’ve partnered with the Docker team to bring over 200+ MCP tools from the official Docker’s MCP catalog to every sandbox. 

Watch the announcement video with Mark Cavage (President & COO at Docker), Tushar Jain (Head of Product & Engineering at Docker), and Vasek Mlejnsky (CEO at E2B).

Model Context Protocol (MCP) is an open-source standard that makes it easy to connect AI systems like AI agents to external applications and use those external applications as LLM tools. In short, MCP tools make AI agents more useful and give them easy access to real world applications.

However, the MCP tools can suffer from various critical security issues that can put your infrastructure at risk. Additionally, having your agents use MCP tools in sandboxes has been until now painful and required a significant amount of sandbox customization making developers spend their time on manually setting up various MCP tools and frameworks inside the sandbox instead of focusing on building the agent.

We’re bringing the MCP support to E2B sandboxes and partnering with Docker to give developers quick and easy access to hundreds of type-safe MCPs in a secure sandbox. The MCPs have been verified and are safe to consume through a single unified interface. What previously took tens of minutes to hours of setup now takes a few seconds.

E2B has already been powering secure sandboxes for agent code paths, and Docker has already produced hundreds of trusted MCP servers. Together, we built a completely safe and trustworthy foundation for connecting AI agents with real-world tools.

This partnership with Docker spans both the content collaboration and the open-source MCP gateway. Together, we’re building on the momentum of a growing ecosystem of open, secure, and interoperable tools—while later introducing commercial extensions on top of it.

The MCP tools in your E2B sandbox can be reached from both inside (accessing the MCP gateway running on localhost) and outside (via the sandbox URL), allowing your local code and sandboxed agents to share the same tools securely.

JavaScript

import Sandbox from 'e2b'
const sandbox = await Sandbox.create({
    mcp: {
        browserbase: {
            apiKey: process.env.BROWSERBASE_API_KEY!,
            geminiApiKey: process.env.GEMINI_API_KEY!,
            projectId: process.env.BROWSERBASE_PROJECT_ID!,
        },
        exa: {
            apiKey: process.env.EXA_API_KEY!,
        },
        notion: {
            internalIntegrationToken: process.env.NOTION_API_KEY!,
        },
    },
});

Python

import os
from e2b import Sandbox

sandbox = await Sandbox.create(
      mcp={            
          "browserbase": {
              "apiKey": os.environ["BROWSERBASE_API_KEY"],
              "geminiApiKey": os.environ["GEMINI_API_KEY"],
              "projectId": os.environ["BROWSERBASE_PROJECT_ID"],
          },
          "exa": {
              "apiKey": os.environ["EXA_API_KEY"],
          },
          "notion": {
              "internalIntegrationToken": os.environ["NOTION_API_KEY"],
          },
      }
)

MCP has become a widely adopted standard for connecting tools to LLMs. By bringing native MCP support to the E2B sandboxes, we’re making it easy for developers and AI agents to securely access all the tools from Docker's MCP catalog such as Browserbase, Stripe, GitHub, Context7, Grafana, or Notion and make their agents more useful in just a few lines of code.

Each MCP tool runs as a Docker container inside the E2B sandbox. The E2B SDK gives you autocomplete, type validation, and one unified interface to control and configure each MCP tool.

How to start

Get started by installing the latest version of the E2B SDK and following quickstart examples in our documentation.

JavaScript

npm install e2b@latest

Python

pip install --upgrade e2b

Towards secure and vetted MCP tools

Our partnership with Docker allows us to bring many official and verified MCP servers in a secure, isolated environment. 

The combination of Docker’s powerful MCP catalog, wide developer adoption, and our secured sandboxes make running MCP servers not only safer but also very easily accessible for all the AI developers building at the forefront of what’s possible with the state-of-the-art models.

We’ve given early access to the sandbox MCP support to selected of E2B users and partners. We want to thank Genspark, Groq, Exa, Browserbase, and others for providing valuable feedback and supporting this launcj.

"E2B MCP with code unlocks an exciting paradigm of MCP in code while keeping the dev experience simple and intuitive."

“Genspark uses E2B Sandbox with MCP integration to let users run 200+ MCPs from Docker’s official catalog with minimal setup, giving access to tools like Notion, Stripe, Browserbase, or Exa. E2B is an essential layer in our infrastructure, enabling us to securely run any MCP directly from GitHub inside isolated sandboxes.”