The EU-hosted AI workspace for teams
Langdock is a German AI workspace that puts several frontier models behind one company login, with EU hosting and GDPR compliance as the selling point rather than an afterthought. Teams get shared chats, shared prompt libraries, their own documents as context, and assistants colleagues can reuse. → Reach for it when the blocker is not "which model" but "legal will not approve a US consumer subscription for the whole department". → Unlike the single-vendor chatbots in Chat with AI, the model is a setting here, not the product.
Not written yet — this chapter is a placeholder for Langdock, one of the all-in-one AI platforms. Research and copy still to come.
Langdock offers a German‑hosted AI workspace where your team accesses multiple frontier models through a single company login, with shared chats, prompt libraries, document context and reusable assistants.
It runs on EU servers and is built for GDPR compliance, so you can meet legal requirements without needing separate US‑based subscriptions for each department.
Yes, the platform treats the model as a configurable setting rather than a fixed product, letting your team choose different frontier models while staying inside the same workspace.
A workspace‑wide repository where pre‑written prompts are stored, categorized, and reusable by any member.
You will have a searchable library of approved prompts that anyone in the workspace can insert into chats or assistants.
A prompt library stores reusable prompts that can be accessed by any chat or agent within the workspace.
You will create a library, add a prompt, and make it available to all members of the workspace.
An assistant (agent) is a custom chatbot configured to answer questions based on specific knowledge bases you provide.
You will upload a document, create an assistant that references it, and test the response.
An AI assistant that answers queries by grounding its responses in your organization’s PDFs and text files.
A live Langdock assistant that references exclusively the uploaded internal knowledge base when responding
A workflow connects multiple Langdock components—such as chats, agents, and integrations—to automate multi‑step processes.
You will design a workflow that sends a user’s question from a chat to the document‑based assistant and returns the answer.
23 outcomes in all — one per recipe below.
Open Langdock, go to the Assistants tab and click “Create New Assistant”. Enter a name, a brief description, and write a concise role prompt that defines the assistant’s task, style, and format. Save it and the platform instantly spins up a live chatbot without any coding.
In the assistant view select the “Knowledge” tab and drag‑and‑drop your PDFs or text files (such as FAQs or product specs) into the folder, or use the upload button. Langdock indexes these files into a vector store, making their content searchable when the assistant answers queries.
Write a prompt that combines role, task, style, and output format—for example: “You are a professional customer service agent for Max Lamp Shop. Analyze incoming inquiries and draft a response. Style: clear, friendly, concise. Format: summary of request, answer, next steps.” Adding extra rules (e.g., always refer to the hotline) ensures consistent replies.
Open the assistant’s chat view and click the model dropdown at the top, then select a model such as Claude 3.5 Sonnet or GPT‑4. Use the Creativity/Temperature slider to set a value (e.g., 0.2 for low creativity) and save; subsequent queries will use those settings.
Need answers from only our internal docs
Langdock lets you create custom assistants by uploading internal documents and configuring system prompts. This works by grounding the LLM in your specific data rather than relying on general training data, ensuring accurate, context-aware responses.
Team members keep reinventing prompts
Instead of each employee writing prompts from scratch, Langdock provides a workspace-wide prompt library. This ensures consistent output quality and preserves institutional knowledge by making proven prompts instantly reusable.
Need to keep corporate AI data in the EU
Langdock's backend allows administrators to toggle which LLM providers are available and filter them by hosting region. This ensures sensitive corporate data stays within EU data centers to meet strict compliance requirements.
Docs scattered in Confluence, Notion and SharePoint
Langdock integrates directly with platforms like Confluence, Notion, SharePoint, and OneDrive. This allows the AI to index and search internal documents natively, solving the common problem of fragmented corporate information.
Langdock lets you select any model from providers like OpenAI, Anthropic, or Google without logging into each service. This keeps your workflow flexible and centralizes prompt management.
Need LinkedIn posts that match my brand voice
Agents are pre‑configured AI chatbots that can be trained with instructions, documents and integrations. By feeding brand guidelines and a content calendar you get instant, on‑brand posts.
Meeting requests land in Gmail
Workflows use a visual node editor to chain together agents, deterministic steps, and external actions. By checking email content, calendar availability, and drafting replies, you get a ready‑to‑send draft for every meeting request.
Want Salesforce info inside your chat
Pre‑built integrations let you pull data from existing SaaS tools directly into Langdock conversations, enabling context‑aware AI assistance without leaving the platform.
Everyone can pick any AI model
Workspace admins can whitelist which AI models are available to users, ensuring that only approved providers are used and helping meet security policies.
Need to gather customer comments on your site
A Form trigger node starts the workflow when someone submits a form. You define required fields like name, feedback text, and optional email, which become the input data for downstream nodes.
Want to pull sentiment, category and a one‑sentence summary
An Agent node runs an LLM (e.g., Sonnet 4) using a custom prompt. By defining a structured output schema, the AI returns predictable fields instead of free‑form text, making downstream automation reliable.
Pull a user’s name and feedback into the prompt
Langdock lets you insert placeholders that pull data from previous nodes directly into an agent’s prompt, ensuring the AI processes the actual user input rather than static text.
Want analysis results sent straight to Slack
An Action node connects Langdock to external services. By referencing both trigger and agent outputs, you can compose a rich Slack message that includes sentiment, category, summary, and original feedback.
Team needs a live feedback form they can fill out
Deployment creates a versioned, enabled instance of the workflow and generates a unique URL for the Form trigger, allowing teammates to submit real feedback that instantly runs the automation.
Want to pick a particular AI model for a query
Langdock lets you either let the autorouter pick the best model or manually select from ChatGPT, Claude, Gemini, Mistral, Llama etc. Selecting a model gives you control over capabilities, cost and latency.
Want a complete research report on any topic
The built‑in web research tool crawls the internet for up to several minutes, aggregates sources and returns a structured summary, saving you manual searching.
Need a branded document fast
By uploading a corporate template, Langdock can fill placeholders with AI‑generated content, producing ready‑to‑use documents while keeping branding consistent.
Langdock can connect to Microsoft services (SharePoint, OneDrive, Outlook, Teams) and search them directly from the chat, returning summaries or full content without leaving the platform.
Need to keep related chats, files and a shared prompt together
Projects let you group related chats, attach files, set a shared system prompt (instructions) and reuse that context across all conversations within the project.
Need a quick customer‑support chatbot that answers from your docs
By using Langdock's built‑in assistant creator you can define name, description and a short instruction prompt, then instantly spin up a functional chatbot without writing code. The platform handles the backend integration and model hosting for you.
Need the AI to answer from our PDFs
Uploading PDFs or text files into Langdock’s knowledge folder automatically indexes them into a vector store, making the content searchable by the assistant at query time.
Need consistent, on‑brand answers from an AI
A well‑structured instruction prompt (role + task + style + output format) guides the LLM to produce consistent, professional replies and ensures it uses the uploaded knowledge correctly.
Need more creative vs deterministic answers
Langdock lets you pick different large language models per assistant and set a temperature (creativity) value, enabling you to balance deterministic answers with more expressive output as needed.
The same set on /recipes, filtered by tool and role.
Ask, share, or report — over on the Heidelberg AI community forum.