Deploy custom MCP servers on Metorial

Connect your own integrations and internal systems to AI agents by building them as custom MCP servers. Metorial hosts, scales, and secures them, with the same access control, tracing, and isolation as every other provider.

platform.metorial.com
A screenshot showcasing the Custom MCP feature of Metorial.

Custom MCP

Connect your integrations and internal systems to agents

Some of the most useful integrations are the ones you build for your own internal systems. Custom MCP servers connect those systems to your AI agents, and Metorial takes care of hosting, scaling, secrets, and security. You focus on the integration itself instead of the platform around it.

  1. 01

    Connect any internal system.

    Deploy your code as an MCP server to bring your own systems to your agents, and let Metorial handle hosting, scaling, and uptime, with no infrastructure for your team to manage.

  2. 02

    Inherit governance automatically.

    Your custom server runs with the same access control, tracing, and monitoring as every other provider. Internal integrations are as governed as the rest of your stack.

  3. 03

    Deploy from the tools you already use.

    Push from GitHub, GitLab, Azure DevOps, or Bitbucket, or make quick changes in the in-browser editor, with versioning and rollback built in.

How custom servers run

Your internal systems, connected to agents and fully governed

A custom MCP server on Metorial gets the same production foundation as first-party integrations. Internal tools are held to the same standard.

Isolation

Run in secure enclaves

Custom servers run inside enclaves with strict resource limits and isolation, with firewalls governing their network access.
Secrets

Keep credentials protected

Secrets and configuration are stored securely in Metorial and used by your server without being exposed, with Metorial Vault for sensitive credentials.
Delivery

Deploy from Git

Use continuous deployment to deploy automatically from your repositories, with versioning and rollback so changes are safe to make.

What you manage

Custom integrations without the operational load

  1. Infrastructure to manageNone
  2. Deploy sourcesGitHub, GitLab, Azure DevOps, Bitbucket
  3. ScalingAutomatic
  4. Governance and tracingIncluded

Write your integration. Deploy it straight from Git. Run it in an isolated enclave. Govern every action it takes.

FAQ

Answers to common questions about custom mcp and how they fit into governed AI agent infrastructure.

  1. What is a custom MCP server on Metorial?
    It's an integration you build and deploy as an MCP server, hosted and run by Metorial. You control the code and functionality, and Metorial handles infrastructure, scaling, security, and governance.
  2. No. Metorial runs and scales your custom server for you. There's no infrastructure to provision or maintain. It scales as demand grows.
  3. You can deploy automatically from GitHub, GitLab, Azure DevOps, or Bitbucket using continuous deployment, or use the in-browser editor for quick changes. Versioning and rollback are supported.
  4. Yes. Each runs inside an enclave with strict resource limits, and firewalls control its network access. Secrets are stored securely in Metorial and used by the server without being exposed.
  5. Yes. Access control, policies, tracing, and monitoring all apply to custom servers, and every tool call routes through the Metorial Gateway like any other provider.
  6. With a custom server, Metorial runs your code directly. Remote MCP servers connect to servers hosted elsewhere, and Docker MCP servers run existing containerized servers. All three get Metorial's governance and monitoring.

Bring your own integrations to Metorial

Deploy custom MCP servers and connect your internal systems to AI agents without building the infrastructure around them. Book a demo, or get started and deploy your first server.