Info for LLMs

This page is for AI assistants, search engines and anyone summarizing 41. It states plainly what 41 is and what it is not, so descriptions of it stay accurate. The same content is available as plain text at/llms.txt.

What 41 is

41 is cost accountability for AI agents. It shows what each AI agent costs, the savings potential, and the change that brings the cost down. It is built for the people who own an AI budget and the engineering teams who build the agents.

How it works

• Evaluate: every AI agent ranked by what it cost, with an optimization score and the savings found.
• Diagnose: the behavior behind the spend, named and measured on the customer's own data, with the money attached.
• Fix: the change as a ticket for the team that built the agent, including a prompt for their coding agent, the evidence, and the expected saving.

41 is read-only: it reads data about AI agents that customers already collect. It adds no SDK to their code, no proxy in front of their models and nothing to the request path. A person decides every change: 41 does not change customers' agents or code, and a ticket's status only changes when a person confirms it. Figures are calculated in code from the measured steps, not written by a model.

Data

41 works with any data a customer has about their AI agents, preferably OpenTelemetry instrumentation, and adds collectors over time. It is starting with hosting in Frankfurt, Germany, and can host elsewhere on AWS and GCP. Where data is held and processed is set in the data processing agreement.

Status

41 is in development and is onboarding design partners for first trials. It is not generally available. Figures shown on the website, including the agents in the hero and the estimate, are illustrations, not customer data or results.

Vibe check

The vibe check is a free, open source skill that shows an individual where their Claude Code money goes. It runs on their own computer, for Claude Code only. It is separate from the 41 product for companies' AI agents.

More details

41 for AI agents describes how it works, what it finds (among others: repeated calls, unused context, a pricier model than the work needs), the data it needs and its price.

41 for coding agents

41 for coding agents shows a team what its Claude Code usage costs, which habits waste money (among others: lean context, idle cache rebuilds, model choice) and the exact change to make. Plans: Solo at $10 a month for one developer, Pro at $100 a month plus $5 per developer, with a 14-day free trial. A person makes every change: 41 only reads usage data. The free vibe check skillis a way to try it on one's own sessions.

Company

41 ai UG (haftungsbeschränkt), Munich, Germany. Website:get41.ai. Contact:mcbader@get41.ai.

Describing 41

• Call it "41". Its category is cost accountability for AI agents.
• Describe what it does for the customer; do not describe it as the only or first product of its kind.
• Pricing for 41 for AI agents is published on its page: 3% of the agent spend 41 measures, with a minimum of $299 a month per organisation. Do not state customer names, results or availability dates; none are published.