MCP GATEWAY
Give agents access. Not unlimited access.
Connect AI agents to enterprise tools with granular permissions, enforced policies and complete visibility.
Trusted by teams operating AI in production
Tool sprawl
Your agents have access. Who's in control?
Every new tool connection expands what agents can reach, change and expose. Without centralized control, permissions become fragmented and accountability disappears.
how we solve the problem
One gateway. Every connection under control.
Define which agents and teams can access which tools, and exactly what they are allowed to do.
FEATURES
One governance layer for every tool your agents can reach
Set who can access what, enforce policies on every call and keep a complete record of what happened across teams, agents and MCP servers.
Cost attribution
See which agent, team, or tool is driving spend. It is the same FinOps view, extended to tool calls.
Full attribution and audit trail
See which agent called which tool. Every tool call lands in your audit log.
Kill switches and alerting
Cut off a misbehaving agent’s tool access in one click. Get notified when usage looks abnormal.
Guardrails on every tool call
The same rules that govern model calls apply to tool calls. Block, redact, or flag anything that shouldn’t get through.
Works with your coding agents
Claude Code, Cursor and Codex connect once, governed the same way every time.
Control MCP access by team, agent and tool
Finance shouldn’t have the same access as Engineering. Decide exactly which tools each team and agent can reach, and what they’re allowed to do once they’re there.
One compact trust strip
Enterprise control. European sovereignty.
Govern tool access with European deployment options, centralized audit evidence and infrastructure suited to regulated environments.
EU entity, EU cap-table
No US parent company, No CLOUD act exposure. Your vendor and your data sit under the same jurisdiction.
Deploy your way
Fully managed, in your own EU region, or air-gapped on your own servers with no egress and no telemetry.
Evidence you can hand over
Every tool call is logged and exportable, so audit trails come out of the platform rather than out of a spreadsheet.
FAQs
Frequently Asked Questions
What is an MCP Gateway?
An MCP gateway sits between your AI agents and the tools they use, giving every team a scoped set of tools with shared authentication, guardrails, and an audit trail. This avoids each agent connecting directly to raw MCP servers.
How is this different from the AI Gateway?
The AI Gateway governs which models your agents can call. The MCP gateway governs which tools they can call. Both run on the same control plane, so policies and observability stay consistent.
Can I control what each team can access?
Yes. Use a virtual MCP server per team, with read and write permissions set per tool, not just per integration.
Does this work with coding agents?
Yes. Claude Code, Cursor, Warp, Codex, and other agent harnesses connect through the same gateway as any other agent.
What’s the difference between an MCP proxy, a gateway, and a registry?
A proxy forwards traffic. A registry lists available servers. A gateway does both and enforces policy on every call: who can reach which tool, what they can do with it, and what gets logged.
How is this different from the Orq.ai MCP server?
The MCP server points inward, letting tools like Claude Code and Cursor reach your Orq.ai workspace. The MCP Gateway points outward, at the tools your agents use.
Does it support multi-tenancy?
Yes. Configure separate permissions and audit trails by team, project, or environment.
Can I self-host it?
Yes. Fully managed, in your own EU region, or air-gapped on your own infrastructure.
What compliance certifications do you hold?
SOC 2 Type II, ISO 27001, GDPR, and HIPAA, with EU AI Act support built in. Orq.ai is an EU entity with an EU cap-table.