Put the latest AI to work, under your control.

Intramura is a managed AI gateway inside your own cloud. Every team gets new models and agent tools as they ship, every request gets an owner and a cost, and your prompts, documents and keys never reach us.

Managed in your cloud from $1,500 a month. No per-request fees. See pricing

The Intramura console's spend view: 7,799.10 dollars spent so far this month across 130,040 AI requests, a month-end forecast of about 98,500 dollars, and spend by team, led by Engineering at 44% and Customer support at 28%.
The spend view in the Intramura console, with illustrative data.

Works with the AI companies you already use.

  • OpenAI
  • Anthropic
  • Google Gemini
  • Google Vertex AI
  • AWS Bedrock
  • Azure OpenAI

Plus Mistral, DeepSeek, xAI, Groq and any other AI company or model with a compatible API, including models you run yourself.

Runs on Kubernetes, AWS ECS or a single Linux machine, in AWS, Google Cloud, Azure, your own servers or with any VPS provider.

Why companies are looking now.

  1. AI spend you can't attribute, or stop

    74% of CFOs say pausing AI usage after an overrun takes days to weeks.

    Deloitte CFO Signals, Q3 2026, reported by TechTarget

  2. Staff already use AI on personal accounts

    Workers at over 90% of companies use personal AI accounts; only 40% of companies have bought official ones.

    MIT NANDA, 2025, reported by Fortune

  3. Every new model is another contract and key

    The average company already uses 11 AI models, and expects 16.

    IBM, 2025, reported by SiliconANGLE

The questions you can answer from day one.

Which AI can each team use?

Approve the models each team and project may use, AI company by AI company. Anything else is blocked. New models appear as soon as AI companies release them, ready for you to approve.

Models for the Support team
  • Claude SonnetAnthropicAllowed
  • GPT-5OpenAIAllowed
  • Gemini ProGoogleBackup only
  • Llama 3Your own serversAllowed
  • DeepSeek R1DeepSeekBlocked

New models appear here as AI companies release them.

Who is spending what, and where does it stop?

Set budgets for the company, teams, people, projects and agents. A hard stop holds a request before it reaches the AI company. A warning lets it through and tells the budget owner.

Budgets this month
  • Whole companyMonthly$31,200 of $50,000Hard stop
  • Customer supportTeam, monthly$8,400 of $12,000Hard stop
  • Data teamTeam, monthly$4,150 of $5,000Warning at 80%
  • Contract review agentAgent, daily$20 of $20Limit reached

Request stopped: the contract review agent reached its daily limit.

Which tools can your agents reach?

Agents reach company tools through one door in your cloud. You choose what each agent may do, and every call is on record.

Tools for the support agent

CRM

  • Read customer recordsOn
  • Update a ticketOn

Payments

  • Issue a refundOff

Email

  • Send a replyOn

Every call goes through your gateway and is on record.

What happened to every request?

Every request leaves a record: who, which model, what it cost and how it went, including the backup that answered when an AI company failed. The prompt itself never reaches us.

One request
Team
Customer support
Person
ID 4f2a
Asked for
GPT-5
Answered by
Claude Sonnet, the backup
Cost
$0.021
Took
1.4 s

Attempts

  1. OpenAIUnavailable
  2. AnthropicAnswered

Prompt and answerNever sent to Intramura

Panels drawn with illustrative data.

And everything around it.

  • One address for every AI company

    Apps and agents call one address in your cloud. Switch AI company without touching them.

  • A backup in an outage

    When an AI company fails, requests move to another one. Staff don't notice.

  • Pay once for repeated questions

    Identical requests are answered from a cache in your cloud, at no extra cost.

  • Recommendations to spend less

    Where a cheaper model or a cache would do the same job, drawn from your own usage.

  • Shared company memory

    Assistants and agents share what your company has taught them, kept in your cloud.

  • A library of agent skills

    Skills are versioned and checked in one place, and reach only the teams you choose.

What never reaches us.

We set up and run your gateway, but we're never in the middle of it.

  • Prompts, documents and keys

    What your staff and agents ask the AI, and what it answers, travel only between your cloud and the AI companies you choose. They never reach us, and nor do the keys to your AI accounts.

  • Your staff list

    Names and emails in your staff list stay in your cloud. We see an ID for each person, and your team still sees names.

  • Our outage is not your outage

    Nothing of ours sits between your staff and the AI. If we go offline, your gateway keeps working.

  • Support only when you allow it

    The access you approve is for running the gateway (installing, upgrading and monitoring it), never for reading requests. For support, you grant time-limited, read-only access.

  • Stored where you are

    Usage records and your console's data are kept in your region.

  • Proof for your auditors

    A full audit log, usage records in the monitoring tools you already run, and a security pack on request.

  • GDPR

    Prompts and your staff list never reach us, so we are not a processor of what your staff send to AI. For the usage records we do receive, we sign a data processing agreement.

  • EU AI Act

    Every AI request leaves a usage record, which you can export to your own monitoring tools and keep.

Questions executives ask.

How is it priced?

From $1,500 a month for a managed gateway in your cloud, by gateway installation and by person who uses AI in the month. There are no per-request, per-token or percentage-of-spend fees, and you keep paying your AI companies and your cloud provider directly. Ask us for a quote.

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. The access you approve is for running the gateway, never for reading requests or answers.

What do we need to have in place?

Somewhere to run it: a cloud account (AWS, Azure or Google Cloud), your own servers or a VPS, and someone who can approve our access. We set it up within your network and security rules and run it from there.

What happens if we stop using Intramura?

The gateway doesn't switch off. It moves to the free plan: it keeps serving your requests, security updates stay available for you to apply, and within the free plan's limits your keys, approved models, routing and a company-wide spending cap keep working. What stops is what you paid us for: us running it, budgets per team and project, single sign-on and staff list sync. Your apps call standard AI APIs at an address in your own cloud, so moving them to another gateway, or straight to the AI companies, is a configuration change, not a rewrite. If you run more than one installation, you choose which one moves to the free plan; the others keep serving on their last settings, without updates. The free plan's user limit applies, and you choose which people keep access.

Find out where your AI budget goes.

Tell us how your teams use AI today and which cloud you run on. We will show you what spend control looks like in your own account.