Jan 16
Loading video player...
Docker has just released its MCP gateway,
which can make agents safer to run and easier to manage.
An “Agent” is a piece of code
that can perform actions on behalf of a user.
Usually, this is a long running process,
that derives all of its intelligence
from an external large language model using API calls.
A major limitation here is that it limits hallucinate,
and can suggest incorrect or dangerous actions to the agent.
Usually, developers try to solve this using
their own validation logic,
but this is time consuming and prone to error.
So Docker has built an MCP gateway, which is like a reverse proxy
with over 200 listed popular MCP servers.
The benefit of this is, as a developer,
we can connect to these MCP servers
without worrying about authentication
or containerization.
All of that is managed centrally by the MCP gateway.
So that if one agent has an issue.
The blast radius is limited, and existing servers
or the remaining agents are not impacted.
You can read more about this on the blog
available on Docker's website.
Thanks for watching.
Cheers!
Docker and E2B can run AI coding agents securely and seamlessly. It's a single trusted repository for over 200 popular MCP servers. This includes GitHub, Perplexity, ElevenLabs, and Stripe. This covers all of the use cases in my startup. As a developer, our agent talks to a single MCP Gateway endpoint. The Gateway handles routing, auth, and policies to every external tool. The major benefit is isolated microVMs. Docker governs how an agent talks to the outside world. If something breaks, the blast radius is limited. Blog: https://www.docker.com/blog/docker-e2b-building-the-future-of-trusted-ai/ #Docker #E2B #MCP