Self-hosted AI platform for companies

AI autonomy for your company.

Feel bigger. Stay in control.

Genai4E gives every team governed AI on your cloud, with your sign-in and your models, and runs the tools they build. Bring what you already have, or start with the open-source stack we provide.

Book a demo See how it works

Your cloud
Your sign-in
Your models

InstallOne server or KubernetesDocker on one machine, or your own cluster. Air-gapped installs work too.

PrivacyZero calls homeNo telemetry. Requests go only to the AI providers you connect.

Models3 + anyOpenAI, Anthropic and Gemini, plus any OpenAI-compatible server.

EvidenceSealed recordsA tamper-evident audit log anyone on your team can verify.

The problem

AI is already in every team. The control isn't.

Keys scattered across teamsOne way in. Keys that follow each person's role.
Surprise invoicesBudgets that stop before they're exceeded.
“Who sent what, where?”Every request recorded, tamper-evident.
Shadow AI tools on laptopsTools built from a template, run inside your walls.

Autonomy by design

Bring your own. Or start with ours.

Use what your company already runs. Where you don't have something yet, Genai4E brings an open-source default. Either way, nothing locks you in.

CloudLive

Run it where you already run things.

Bring yours

AWSMicrosoft AzureGoogle CloudYour data centre

Use ours

One Linux server with Docker, or your Kubernetes cluster. Our installer sets everything up, even air-gapped.

Sign-inNew

Sign in the way your company already does.

Bring yours

Microsoft Entra IDOktaAny OpenID Connect

Use ours

Keycloak, open source, set up with Genai4E. One sign-in for the platform and its Git server.

Live on the Kubernetes install; the single-server install follows. Admins keep a password for emergencies.

AI modelsLive

Your enterprise models, or open ones.

Bring yours

Azure OpenAIOpenAIAnthropicGoogle GeminiYour own endpoint

Use ours

Open-source models such as Llama, Mistral or Qwen on your own GPUs, through vLLM or Ollama.

The route

Seven stops between a team's first request and your auditor's report.

Teams see a simple product. You see the whole company, on one page.

01Stop 01 · Teams

Start in minutes.

A team creates a project, picks the approved models, sets a monthly budget and invites people. Each person uses their own key, with code that works with OpenAI's SDKs.

Nothing new to learn. Nothing to install on their side.

Acme / New projectStep 2 of 4

✓Name
2Models
3Budget
4Team

Which models may Support bot use?

Only models your admin has approved are listed.

✓GPT-4.1 mini · OpenAIBest for Everyday chat and short summaries≈ $0.001
per request

✓Claude Sonnet 4 · AnthropicBest for Long documents and careful writing≈ $0.006
per request

Gemini 2.5 Flash · GoogleBest for Fast, high-volume tasks≈ $0.0008
per request

Priya's key, narrowed to Support botPython · OpenAI SDK

from openai import OpenAI

client = OpenAI(
  base_url="https://ai.acme.internal/v1/projects/support-bot",
  api_key="gk_7f3a…",
)

02Stop 02 · Models

Every model, one way in.

OpenAI, Anthropic and Google Gemini, plus any OpenAI-compatible server. Admins choose which models each project may use.

Smart models: back-up models take over when a provider is down, and split traffic shares requests between models.

Acme / Models / support-defaultSmart model

Back-up modelsSplit traffic

If a model fails, the next one in line answers. Nobody changes their code.

1OGPT-4.1OpenAI · primaryDown · 2 min

2AClaude Sonnet 4Anthropic · back-upAnswering

3YLlama 3.3 70BYour servers · vLLMReady

Answered by Claude after GPT failedreq_8f21 · 2.4 s

03Stop 03 · Cost

Spend you can see and cap.

Cost per project, model, team and person. Monthly budgets, an alert at 80% and an optional hard stop.

Rate limits per project and per model. Chargeback by team, ready for finance.

Acme / Usage & CostOct 1 – 9, 2026

Spent so far$4,812

Forecast$6,950

Budget$8,000

Support bot

$1,240 / $2,000

Finance AI

$860 / $1,000

Data science

$1,410 / $3,000

Contract review

$312 / $500

Finance AI is at 86% of its $1,000 budget. Sam was alerted at 80%. Requests stop at 100%.

04Stop 04 · Activity

See what happened.

Every request appears in Activity, step by step: request, model, tools, answer. With the time, tokens, cost, and feedback from the person who asked.

Privacy is set per project: keep full text, numbers only, or redacted.

Support bot / Activity / req_8f21Oct 9 · 10:42

Answered“Summarise ticket #4821 for the customer”

Time2.4 s

Tokens1,284

Cost$0.0061

Asked byPriya

Request receivedFrom Support bot12 ms

Checked key and budgetWithin budget · 62% used4 ms

Model · Claude Sonnet 4Decided to look up the ticket0.43 s

Tool · search_ticketsHelpdesk server · 3 results0.41 s

Model · Claude Sonnet 4Wrote the answer · 182 words1.49 s

Privacy · Keep full textFeedback from Priya · Helpful

05Stop 05 · Audit

Proof for your auditors.

Every change and access sealed in a tamper-evident audit log that anyone can verify.

A compliance report per project, an AI inventory and an access review, all exportable.

Finance AI / Compliance reportExported Oct 9, 2026

Project

Finance AI

Seal verified ✓

Purpose: Summarise supplier invoices and flag unusual amounts.
Risk level: Limited · reviewed by Sam, Sep 30
Data: Financial recordsSupplier contacts
Members: Priya, Sam and 4 others

sha256 9c4e 1b07 52aa … e310 a71b
2,418 audit entries · chain intact

06Stop 06 · Tools

Tools your AI can use.

A catalog of approved MCP tool servers. See who owns each one, what it can do, and whether it's healthy.

Projects only reach the tools they've been given.

Acme / Tool catalog3 approved servers

Helpdesk

Owned by Support

search_ticketscreate_ticket

Healthy · 120 ms

Finance ledger

Owned by Finance AI

get_invoicelist_suppliers

Healthy · 95 ms

Internal docs

Owned by Platform

search_docsread_page

Slow · 1.4 s

07Stop 07 · Hosting

Runs inside your walls.

Installs on your own server or your Kubernetes cluster. Works air-gapped. No telemetry, no phone-home, provider keys encrypted.

Every app runs in its own fenced space, with locked-down containers and no cluster access for anyone's code.

Your network

Teams and apps

Support bot, Finance AI, your MCP servers

Genai4E

Keys, budgets, limits, records, sign-in

Your own models

vLLM or Ollama on your GPUs

Audit logEncrypted provider keysLocked-down containersNo telemetry

$ ./genai4e init && ./genai4e up

OptionalOOpenAIAAnthropicGGemini

Model requests only. Nothing else leaves your network.

New · Apps

From a template to a running tool, by git push.

A project can now be an app that Genai4E builds and runs for you. Start with an MCP server: the tools your coding agents and assistants use, governed like everything else.

MCP servers New
Knowledge search (RAG) Next
Agents Next

01CreateChoose “MCP server” when you create a project.
02CodeA Git repository appears with a FastMCP template, tests and instructions for coding agents.
03PushEvery push to main is tested, built and published. No Docker, no special access.
04RunGenai4E deploys it into its own fenced space, with its own key, so budgets, limits and audit apply.
05ConnectCopy one command into Claude Code. Everyone connects with their own key.

support-tools — Priya's laptop

$git commit -am "Add a lookup_order tool" && git push
✓tests passed · 4 tests
✓image published · p-support-tools/support-tools:4f2a91c
✓version #3 live · the previous one kept running until it was healthy
$claude mcp add --transport http support-tools \
https://ai.acme.internal/apps/support-tools/mcp \
--header "Authorization: Bearer gk_9c1e…"
✓Added MCP server support-tools
$

Its own key

Its AI calls count against the project's budget and show up in Activity.

Its own space

Locked-down containers. Deny-by-default network rules keep it apart from your other apps.

No cluster access

CI asks, Genai4E deploys. A failed version is rolled back on its own.

The road ahead

Built for what your company does next.

Live

Runs on AWS, Azure, Google Cloud or your own servers
AI models for every team, with keys that follow people's roles
Budgets and limits
Activity for every request
Audit and compliance
Tool catalog

New

Single sign-on: Keycloak (open source) or your own provider
MCP apps from a template, built and run on your Kubernetes
Git, CI and an image registry built in

Sharing apps between teams
Knowledge search (RAG) and agent apps

Later

A Helm chart for EKS, AKS, GKE and on-prem clusters
Bring your own GitLab
A shared prompt library

One road for everything your company builds with AI.

Book a demo

Say yes to AI. Keep control.

A 30-minute call with an engineer. We'll show Genai4E running on a server like yours, with your cloud, your sign-in and your models, and answer your security team's questions.

Prefer email? hello@genai4e.com

AI autonomy for your company.AI autonomyfor your company.

AI is already in every team. The control isn't.

Bring your own. Or start with ours.

Run it where you already run things.

Sign in the way your company already does.

Your enterprise models, or open ones.

Seven stops between a team's first request and your auditor's report.

Start in minutes.

Every model, one way in.

Spend you can see and cap.

See what happened.

Proof for your auditors.

Tools your AI can use.

Runs inside your walls.

From a template to a running tool, by git push.

Built for what your company does next.

Say yes to AI. Keep control.

AI autonomy for your company.