In Development

Citadel

Self-hosted AI proxy. See everything your organization sends to every AI provider.

An npm package and standalone executable that intercepts AI requests, captures telemetry, and feeds a dashboard you own. Your infrastructure, your data, complete visibility.

terminal
# install globally
$ npm install -g @majorlinkx/citadel
# start the proxy
$ citadel start

6+

AI Providers

100%

Self-Hosted

0

Data Shared

<5min

Setup Time

Architecture

How Citadel Works

A proxy on your machine, an API on your infrastructure, a dashboard in your browser.

Local Machine
npm package / executable
Citadel Proxy
intercepts all requests
AI Providers
OpenAI, Claude, Gemini...
telemetry
Citadel API
self-hosted on your infra
PostgreSQL
your data, your schema
Dashboard UI
roles, costs, analytics
Setup

Configure in Minutes

Point Citadel at your providers, set your API endpoint, and start capturing telemetry. One config file. No vendor calls, no procurement cycles.

1

Install the package

npm install globally or download the standalone executable.

2

Add your providers

Point Citadel at OpenAI, Anthropic, Google, Bedrock, or any upstream.

3

Set the API endpoint

Telemetry flows to a Citadel API instance on your infrastructure.

4

Open the dashboard

Costs, usage, user roles, and audit trails from day one.

citadel.config.yaml
# citadel.config.yaml
proxy:
  port: 8080
  providers:
    - name: openai
      upstream: https://api.openai.com
    - name: anthropic
      upstream: https://api.anthropic.com
    - name: google
      upstream: https://generativelanguage.googleapis.com

telemetry:
  endpoint: https://citadel-api.internal
  batch_size: 100
  flush_interval: 30s

dashboard:
  port: 3000
  auth: oidc
Features

Know What You Are Spending on AI

Built for organizations that take AI seriously enough to track it properly.

Self-Hosted AI Proxy

npm package or standalone executable

Citadel runs on your machine as a proxy between your team and every AI provider. Every request routes through it, giving you a single point of visibility without changing how anyone uses AI.

Cost Center Visibility

know what you are spending and where

How much is your company spending on AI? Which teams drive the cost? Which models burn budget without delivering value? Citadel answers with real data: breakdowns by team, project, model, provider.

Infrastructure Audit

full request and response visibility

Every request is logged with a full audit trail. See what data is being sent to which provider, when, and by whom. Since Citadel is a proxy, nothing bypasses it.

Dashboard with Roles

user management, CRUD, analytics

A built-in dashboard served from the Citadel API. User roles, team management, cost center analytics, and budget enforcement. No separate BI tool required.

Supports Every Provider

Citadel is not locked to a single vendor. It proxies requests to any AI provider, including local models running on your own hardware.

Deploy as Docker, Kubernetes, or bare metal. Citadel adapts to your security requirements, not the other way around.

AI Providers

OpenAI / ChatGPT
Anthropic / Claude
Google / Gemini
AWS Bedrock
Azure OpenAI
Local Models

Deployment Options

Docker
Kubernetes
Bare Metal
Cloud VMs

Take control of your AI spending

Citadel is currently in early access. Get in touch to deploy unified AI cost visibility on your own infrastructure.

Request Early Access