> For the complete documentation index, see [llms.txt](https://wiki.dieagenturverwaltung.de/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://wiki.dieagenturverwaltung.de/die-agenturverwaltung-en/einstellungen-1/integrationen/mcp-model-context-protocol.md). # MCP (Model Context Protocol) MCP (Model Context Protocol) is an open standard that allows AI models like Claude to connect uniformly with external tools, data sources, and services, instead of building a separate custom solution for each integration. Your Die Agenturverwaltung provides its own MCP server, allowing Claude to access its data and functions directly and use them in conversations. To connect a language model with MCP, first create in Die Agenturverwaltung under *Settings > Integrations > Model Context* Protocol a token.\ This is a unique URL that you can later use as a connector in an application like Claude, ChatGPT, Langdock, or InnoGPT. {% hint style="info" %} **Warning: Anyone who has access to this token also has access to the data in Die Agenturverwaltung that you have access to.**\ Do not store this token unencrypted on your computer and only pass it on to trusted people. {% endhint %} {% @arcade/embed flowId="Rx0XpGhaisOUtN3Vl3Gp" url="" %} ## What can I do with MCP? You can give a language model - e.g. Claude, ChatGPT, Mistral, Kimi - access to the data in Die Agenturverwaltung and have conversations with this model about your data.\ \ Here are a few **example questions**, that you can ask an LLM chat: #### **Revenue & result** * What was the billed revenue in H1 2026, broken down by customers/customer accounts? * Show me the gross profit (contribution margin) per customer account for Q2 — absolute and as a percentage of revenue. * What is the result after external services and internal staff costs per job? * Revenue by service period vs. by invoice date — where do the figures diverge? #### **Liquidity & receivables** * Which sent invoices are still unpaid, sorted by invoice date? * How much money actually came in in July, how much was invoiced? * Which credit invoices/cancellations were there this year and for which customers? * What is the prepayment volume and how much of it has already been invoiced? #### Pipeline & forecast * How high is the approved estimate volume of the last 90 days? * What estimates are awaiting approval — weighted by probability? * Which budgets from approved estimates have not yet been billed? #### **Productivity & cost structure** * What is the average hourly rate of billed agency services per customer account — and where is it below plan? * What is the freelancer share of agency services, broken down by trade? * Gross profit per head based on internal staff costs, by team? * Billable vs. total working time — what is the billability? #### Project control / risks * Which jobs have exceeded their agency service budget and by how much? * Where has the budget been used up by >80 %, but the service period is still running? * Which top 10 customers account for what share of revenue?” (concentration risk) #### Procurement side * Which purchase invoices/supplier orders accrued in period X, by job? ## Which language models you should use {% hint style="info" %} The quality of the answers depends very strongly on the selected language model. **Do not use overly economical models** like Haiku or o3 mini - especially when the language model is supposed to relate multiple lists, e.g. users and metrics, e.g. hours incurred, to each other and independently decide which tools it should use. In most cases, these models cannot keep track of the available tools and already fail on simple questions.\ .\ **We recommend e.g. Sonnet 5, Opus 4.7 and larger, ChatGPT Earth or Sol.** {% endhint %} ## Privacy notice on working with external service providers Make sure that you can also use services such as Claude or ChatGPT in your company in compliance with GDPR; for this you need an enterprise account and must make certain presettings, e.g. the hosting of the language models within the EU. First clarify the use with your data protection officer, as the MCP can also be used to read out customer- and person-related data. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://wiki.dieagenturverwaltung.de/die-agenturverwaltung-en/einstellungen-1/integrationen/mcp-model-context-protocol.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.