Imagine: You’ve done it. After weeks of work, you’ve rolled out your own private AI platform for your company based on LibreChat or OpenWebUI. Your colleagues are excited about the new possibilities. But after the first month, the finance department and CEO come knocking. The question isn’t even about the size of the bill, but much more fundamental: “What exactly is AI bringing us?” Suddenly it becomes clear: You’re investing in AI but can’t clearly quantify the ROI of these initiatives. Your platform is a black box that makes it impossible to evaluate which AI applications are delivering real business value and where you should increase your investments. This scenario is the nightmare of every IT manager and the reason why many ChatGPT for enterprise projects stall. Without detailed insights into usage, cost control is impossible and demonstrating ROI becomes pure guesswork.
At innFactory AI Consulting, we faced exactly this challenge. That’s why we, together with our parent company innFactory GmbH, developed a solution. Today we give it back to the community: the LibreChat Admin Dashboard, our new open-source project that is also part of our AI stack within CompanyGPT.
The Problem: Lack of Insights into LibreChat
LibreChat is an excellent foundation for building a private AI infrastructure. However, it lacks a crucial feature out of the box: an admin interface for analyzing usage. How many users are using the private GPT? Which AI agents and MCP tools are being used? How many searches and file analyses are being handled by the system? And the most important question of all: Which queries and models are driving the costs for token billing?
Without this GPT analysis, companies are navigating blind or only seeing a very superficial analysis through their Azure billing at the end of the month. This is exactly where we step in with our Admin Dashboard, providing the urgently needed instruments.
Full Transparency with the LibreChat Admin Dashboard
Our dashboard is a lean but powerful analysis tool that directly accesses the database of your LibreChat instance. Published under the liberal MIT license on GitHub and immediately deployable via Docker, it provides you with all the key metrics at a glance. As part of our CompanyGPT stack, this tool is now included by default if you wish.
Let’s look at what insights the dashboard provides in practice:
The LibreChat Admin Dashboard delivers all important metrics for usage analysis at a glance.
The dashboard gives you the tools to understand how your AI platform is being used:
- Usage KPIs at a glance: You can immediately see that in the selected time period, 6 of the total 21 users were active and made 185 queries in 48 conversations. The comparison to the previous period is particularly insightful: User numbers have decreased, but the number of conversations and processed files has increased significantly (+17 and +24 respectively). This indicates that the remaining users are using the platform more intensively and for more complex tasks.
- Detailed token analysis: The dashboard breaks down the token metrics precisely. In the example, 1.01 million input tokens were consumed, but only about 105,000 output tokens were generated. This is a strong indicator that users are passing very long documents or complex instructions to the AI (high input), but receiving concise, short answers (low output) – a typical pattern for analysis and research tasks.
- Insights into tool usage: The “MCP Calls” and “Web Searches” metrics are crucial for understanding the usage of connected tools (agent tools). A strong increase in MCP calls (+13) shows that your users are learning to leverage the true strengths of AI: interaction with other systems.
- The truth lies in the agents: The “AI Agent Statistics” table is the heart of the analysis. Here you can see in black and white which agents are providing the most value. In the example, the “Research” and “CompanyGPT Doku” agents dominate with 37 and 35 calls respectively. Others, like the “Meeting Protokolle Agent,” are hardly used. This is priceless information! You can now specifically promote popular agents, offer training for less-used tools, or question their functionality.
An Analysis Tool, Not a Billing Tool
An important note: The dashboard serves analysis and monitoring purposes, not final token billing. The exact cost data still comes directly from your cloud provider (e.g., from Azure Cost Management). Our tool accesses the transaction data in the LibreChat database. It’s your early warning system to identify cost drivers and optimize usage long before the bill surprises you.
A Central Building Block of Our CompanyGPT Architecture
For us, this dashboard is more than just a useful tool. It’s a fundamental building block of our CompanyGPT AI architecture. With CompanyGPT, we offer companies a turnkey, GDPR-compliant, and secure AI platform that operates in the customer’s own cloud infrastructure. Transparency and control are not optional add-ons but are at the core of our promise. The dashboard is the logical next step to underpin this promise.
A Look at Practice: Our Own AI Usage Data at innFactory
But enough theory. To show that this is more than just another tool, we live this transparency at innFactory ourselves. The question of whether new AI tools like our CompanyGPT or GitHub Copilot are really being adopted in the company is crucial for success. A look at the real usage data of our own team in Q4 2025 provides a clear and impressive answer:
Real insights: The usage statistics of innFactory GmbH for Q4 2025 show a massive increase in AI adoption.
The numbers for our 30 active users compared to the previous quarter (Q3) speak for themselves:
- Queries: +4,833 to over 16,000
- Conversations: +365 to almost 1,900
- Processed input tokens: +74.65 million to over 207 million
This massive increase, especially in input tokens, shows us two things: First, employees are not only using the AI tools more frequently but also for increasingly complex tasks. Second, the combination of providing the tools and targeted training leads to sustainable and deep integration into daily work. Internally, the use of our CompanyGPT is probably only surpassed by GitHub Copilot in software development. This data-driven insight allows us to make success measurable and to strategically control future investments in AI competencies.
Take Back Control of Your AI Platform!
We have now firmly integrated the LibreChat Admin Dashboard into our CompanyGPT stack.
For our existing CompanyGPT customers: Contact us! You can benefit from this new analysis tool immediately and gain unprecedented transparency about the usage and costs of your AI platform. As part of monthly maintenance, there are no additional costs for activation.
Are you looking for a secure, transparent, and powerful AI solution for your company? Then now is the right time for a conversation. Visit company-gpt.com to learn more about our approach and how we can accompany you on the path to a data-driven AI strategy.
And for all developers and tinkerers: Visit our project on GitHub. We look forward to your feedback, ideas, and contributions.