Connect every tool
Give agents access to 1,100+ MCP servers and 20,000+ tools, all from one place.
Put AI to work across your company, safely
Every agent gets approved MCP tools and skills, tied to the person behind it. One MCP gateway, governed by role and recorded call by call.
The problem
AI reaches company tools faster than anyone can govern it
The moment a second team wants agents connected to real apps, the workarounds start. These are the three that stall most rollouts.
- 01
Shadow MCP
People install MCP servers on their own laptops to get work done, and nobody knows which tools reach company data. What is shadow AI?
- 02
API keys in config files
A personal token pasted into a local config works, so a teammate copies it. Months later it still works, with no owner and no expiry. Are MCP servers secure?
- 03
No record of what ran
When an agent changes the wrong record, the first question is whose access it used. A shared credential has no answer. Audit agent activity
Connect, govern, and see
every agent call.
Metorial puts one governed layer between the AI clients your company uses and the tools they reach. See how the MCP gateway works.
Connect, govern, and see
every agent call.
Metorial puts one governed layer between the AI clients your company uses and the tools they reach. See how the MCP gateway works.
Check every agent action
Protoguard checks every message and tool call against your policies before it reaches your systems.
Add your own tools
Connect internal systems and private APIs as your own providers.
Metorial Workforce
A home for your team's agents
People sign in, see the tools and skills they're allowed to use, and connect their agent with one link.
Sign in with your SSO
Decide who can use which agents, tools, and providers, using the logins you already have.
Keep your data safe
Everything's encrypted, every integration is isolated, and you can run Metorial in your own cloud.
Agent Skills
Build skills together
Your experts write a workflow once, and every teammate's agent can use it.
How Metorial works
One governed path from the AI client each person already uses to the company tools they are allowed to reach.
- 01
Sign in once with your SSO
People sign in to their portal with the identity provider you already run. Their groups decide which integrations and skills they see.
- 02
Paste one URL into any client
Each person adds their Magic MCP URL to Claude, ChatGPT, Cursor, Codex, or their own agent. The same approved tools show up everywhere.
- 03
Every call is checked and logged
Each tool call is checked against policy, runs on that person's own connected account, and is recorded with their identity.
People sign in, see the tools and skills they're allowed to use, and connect their agent with one link. You stay in control of who can do what.
Portals
One place to connect your team
People sign in, find the integrations and skills they’re allowed to use, and connect their agent with one link. You control who gets access to what.
Skills
Share how your team gets work done
Write a skill once and share it with the people who need it. Instructions and tools stay together, so agents follow your team’s approach.
Agents
Put your tools to work in any agent
Connect the agent your team already uses to approved tools and shared skills. Metorial handles access while people keep working in a familiar interface.
How every tool call stays secure
The same four checks apply to every agent and every client, from the first sign-in to the record your security team reviews.
- 1
Sign in with your identity provider
People sign in with SAML SSO, and SCIM keeps accounts and groups in step, so leavers lose access when they leave. - 2
Policy decides what is allowed
Access policies grant tools by person and group, down to single tools and read-only sets. - 3
Protoguard screens the call
Protoguard checks every request for prompt injection and watches providers for unexpected schema changes. - 4
Frequently asked questions
The questions teams ask most before rolling Metorial out.
What is Metorial?
Metorial is an open-source MCP and skills gateway. It connects AI agents and assistants to the tools and data your company uses, behind one sign-in, with access control and an audit log. See how the MCP gateway works.What is an MCP gateway?
An MCP gateway sits between AI clients like Claude, ChatGPT and Cursor and the MCP servers that connect them to company apps. It handles routing, sign-in, access rules and logging in one place, so each team does not solve them again.Which AI clients does Metorial work with?
Any MCP-compatible client, including Claude, ChatGPT, Cursor and Codex. Each person pastes one Magic MCP URL and gets the same approved tools, so switching client does not mean setting access up again.Can we self-host Metorial?
Yes. Metorial is open core on GitHub and runs self-hosted or on-prem, or you can use Metorial Cloud. See how it works.How does access control work with our identity provider?
People sign in with the SSO you already run, and SCIM keeps accounts in step. Admins grant tools by role, and access control applies everywhere at once, across every client and agent.Is Metorial secure and compliant?
Every tool call runs on a real person's identity and is recorded. Metorial is SOC 2 Type II and GDPR compliant, with EU and US regions, and checks each call for prompt injection. See security.How much does Metorial cost?
There is a free Dev plan, a fixed-price Scale plan, and custom Enterprise pricing. See pricing for the current plans.Can we add our own MCP servers?
Yes. Alongside the catalog, you can connect your own internal servers and remote MCP servers from other vendors, behind the same access rules and audit log.