Docker + E2B: Building the Future of Trusted AI

Trusted Software Starts Here

The era of agents is here. Some teams are experimenting, others are just getting started, and a few are already running agents in production. But one challenge stands out: trust. Trust that your agents will act securely. Over 20 million developers already rely on Docker to build and ship software safely and fast. Now, we’re helping you build what’s next.

Over the past few months, Docker has continued to make significant investments in driving developer productivity and building a trusted foundation for AI, with a focus on simplicity and portability, from how you build and run agents to how you secure them. We’ve shipped new capabilities: a toolkit for MCP, support for running open-weight models locally, and a catalog of 200+ MCP servers. 

Today, we’re taking that commitment to the next level through a new partnership with E2B, a company that provides secure cloud sandboxes for AI agents. Together, we’re giving developers fast, secure access to hundreds of real-world tools, without sacrificing safety or speed.

The Next Frontier of AI: Trust

Agents can code, connect, and act, but without control, that power creates risk in two areas.

First, when agents run code. AI-generated code often executes in live environments without safeguards, accessing files or APIs it shouldn’t. Teams are learning that “auto-run” doesn’t mean “safe to run.”

Second, when agents connect to real-world tools. Integrating with external tools can create security trade-offs. And the challenge keeps growing. With the rise of the Model Context Protocol (MCP), every connected tool becomes its own potential risk surface with hundreds of servers and tokens to manage. Today, developers use a variety of DIY isolation techniques, custom-built sandboxes, and resource limits, but all of them tend to slow productivity.

Developers have seen this before. Before containers, developers struggled with messy dependencies and inconsistent environments when building applications. 

Today’s AI builders face a similar challenge. Models need access to APIs and databases, but enabling that access slows them down and introduces complexity and risk. Just as containers standardized how applications run, Docker is now standardizing how agents connect and act.

This is where E2B and Docker come together. E2B secures agent-generated code execution with cloud sandboxes. Docker secures tool access through the MCP Gateway and Catalog, offering direct connectivity to 200+ real-world tools. Each MCP server is curated by Docker and automatically audited for exploits and malicious behavior.

Docker + E2B: Secure Access to Hundreds of MCP Tools

Starting today, every E2B sandbox includes direct access to Docker’s MCP Catalog, a collection of 200+ tools such as GitHub, Perplexity, Browserbase, and ElevenLabs, all enabled by the Docker MCP Gateway. Developers can now run and connect agents more confidently, without slowing down.

Try it here.

What’s Next 

This is only the beginning. Today’s partnership is the first step in a larger journey. Together, our goal is to build the future of a secure AI stack, where every agent runs securely, every connection is verifiable, and developers can move fast without compromise.

From Docker’s perspective, our goal is to ensure developers can trust the MCP servers they pull, validate them through Docker, and govern their AI stack with the same visibility and confidence they already have for containers today.

The next wave of AI development will be built on trust. Trust in the tools, the data, and the environments behind them. At Docker, we’re building that foundation.