MCP built in: one governed gateway for every tool your agents touch.
Windows only for now.
drop a screenshot at assets/gateway.png(MCP Gateway page)
drop a screenshot at assets/gateway-add.png(Attach MCP server form)
The Model Context Protocol is the open standard for agent tools, and mimOE speaks it at the platform level. Anything with an MCP server (Jira, databases, browsers, your own scripts) becomes an agent tool.
Agents never connect to tool servers directly. Servers are attached to the MCP gateway once, as a remote endpoint or a local command, and the gateway brokers every call.
Each agent authenticates with its own key, and a key only reaches the servers it was granted. Revoke the key, and that agent's tools are gone. No redeploy, no hunting through configs.
The shell your agents use is itself an MCP server behind the gateway, governed by the same per-agent policy as every other tool.
Because every tool call crosses one gateway, there is one place to see, allow, and audit what agents do with the outside world. It's the same choke point Trace and Governance build on.
Point the gateway at a remote MCP endpoint, or give it a local command with args and env, and the gateway boots and supervises it.
Create client keys and grant each one specific servers. A key is the only way in, and it opens only the doors it was given.
Deploying an agent registers its shell with the gateway and hands the agent a scoped key. Tools work out of the box.
At run time each tool call flows through the gateway: authenticated, routed to the right server, and visible in one place instead of scattered across processes.