Your data leaves
Prompts carry source code, customer records and internal documents. A hosted gateway becomes one more place they are processed and logged.
A managed AI gateway, in your cloud
Your AI gateway,
within your walls.
Intramura deploys and operates an open source AI gateway inside your own cloud account. Prompts, keys and logs stay with you. We handle the install, the upgrades and the pager.
- Runs in your AWS, Azure or Google Cloud account
- Built on open source, no lock-in
- Wired into the tools you already run
The problem
A hosted gateway sees everything your AI does.
Every prompt, every response and every provider key passes through the gateway. When someone else hosts it, that traffic sits on their infrastructure.
Your keys leave
Provider credentials for every model you use are stored with a third party, outside your own secrets management.
Running it yourself is a job
Self-hosting fixes both, but someone has to install it, upgrade it, watch it and connect it to everything else. That is the part we take on.
How it works
The gateway lives in your account. We operate it.
The gateway runs inside your cloud account. Requests go from your apps to the models you choose, and nowhere else.
Intramura
Deploys, upgrades, monitors
Operations access only
Your cloud account
Callers
Your apps and agents
Intramura platform
AI gateway
Routing, limits, keys, usage
Models
Self-hosted models
Telemetry to
Your observability stack
Teams and budgets from
Your HR and identity systems
Outside, by your choice
Hosted model providers
Only the ones you enable
-
We deploy
We install the gateway into your account, sized for your traffic and set up to your network and security rules.
-
We integrate
We connect it to your identity provider, observability stack and HR system, so teams, budgets and dashboards match how your organisation works.
-
We operate
We handle upgrades, monitoring and incidents. Your platform team keeps control of policy without carrying the day-to-day.
Platform
A complete AI gateway, without running it yourself.
What the platform does
- One API for every model. A single OpenAI-compatible endpoint for hosted providers and self-hosted models.
- Routing and failover. Send each request to the right model, and fall back when a provider is down.
- Spend attribution. Token usage tracked by team and application, so costs have an owner.
- Quotas and rate limits. Token-aware limits that stop one workload from draining the budget.
- Central credentials. Provider keys held in one place, so application teams never handle them.
- Tools for agents. Controlled access to MCP servers, the standard way agents reach tools and data.
What we take care of
- Deployment in your account. Installed to your standards, in your network, under your controls.
- Upgrades. New gateway releases tested and rolled out for you.
- Monitoring and response. We watch the gateway and act when something breaks.
- Custom integrations. Connections to your observability and HR systems, built for your setup.
- A team that knows cloud. Years of experience running production infrastructure across the major clouds.
Why Intramura
A gateway that fits your organisation, not the other way round.
An AI gateway is only as useful as its fit with the systems around it. The difference is who runs it, and how well it works with what you already have.
Years of cloud expertise
We have spent years building and operating infrastructure in enterprise cloud environments. Your networking, security and compliance constraints are our starting point.
Observability, your way
Gateway metrics, logs and traces flow into the observability stack your teams already use, in the shape your dashboards and alerts expect.
Mapped to your org chart
We integrate with your HR system so teams, cost centres and access follow your real organisation, and stay current as people join, move and leave.
Where it runs
Your cloud, your account, your region.
AWS
Microsoft Azure
Google Cloud
Deployed in the regions you choose.
Open source underneath
No lock-in, by design.
The gateway is built on an open source, Envoy-based project under the Apache 2.0 licence. What runs in your account is that open source software, not a proprietary fork. If you ever stop working with us, it keeps running and it stays yours.
FAQ
Common questions
Do my prompts ever reach Intramura?
No. The gateway runs in your cloud account, and requests travel from your applications to the model providers you enable. Intramura's access is for operating the gateway, not for reading its traffic.
What is an AI gateway?
It is the single point your applications and agents go through to reach AI models. It gives applications one API for many models and gives platform teams one place for credentials, routing, quotas and usage.
What do we need to have in place?
A cloud account on AWS, Azure or Google Cloud and a team that owns it. We work with you on the access we need and deploy within your network and security rules.
Which observability and HR systems do you integrate with?
The ones you run. Integrations are built for each customer, so tell us what you use and we will scope it during the demo.
What happens if we stop using Intramura?
The gateway is open source and runs in your account, so it keeps working. You take over operating it, or hand it to another team.
Which open source gateway do you use?
We are glad to walk through the full stack, including the project and its licence, as part of the demo and any security review.
Bring your AI gateway inside the walls.
Tell us about your cloud and the systems you need it to work with. We will show you what a deployment looks like.