Build a chatbot in minutes. For your website.
Point Ragety at your website or drop in a folder of documents. It reads them, and whoever asks gets an answer in plain language, with the passage it came from. No vector database, no project plan, no engineers.
Ragety is at an early stage. We onboard a limited number of companies whose use cases feed into development.
What is Ragety?
Your people are already using AI. Mostly where you cannot see it.
In most companies the question is no longer whether AI arrives, but whose rules it arrives under.
-
Company documents in private accounts
Staff paste contracts, tickets and customer mails into consumer AI accounts, outside any processing agreement, because it is the fastest way to get an answer. -
No approved route for sensitive content
Anything under a duty of confidentiality has nowhere sanctioned to go, so the departments with the most to gain are the ones told to wait. -
The pressure for speed stays
Management asks for the productivity that strong models and assistants promise, and IT is expected to deliver it without giving up control of the data.
One knowledge base. Wherever the question comes up.
You connect your content once. The same answers, the same permissions and the same record of every request, on every surface already in use in your company.
Your website
availableInternal knowledge portal
availableSlack
availableChatGPT, Claude and others
plannedAnd behind every answer, the model that fits the task
Answers run on models via AWS Bedrock or on models we host ourselves, and a switch changes nothing in your integration. What you can choose from today:
OpenAI
Anthropic
Meta Llama
Mistral AI
Amazon Nova
Self-hosted
Five clicks, and the first answer is on the screen.
This is the entire setup. The only thing you bring is the content, and someone who can tell whether an answer is correct.
-
1
Create a project
Give it a name and pick who it is for. That is the whole form.
Click 1 of 5 · New project
-
2
Drop in your documents
PDF, Word, PowerPoint, HTML, Markdown and plain text, individually or in bulk. Or paste your sitemap and let Ragety read the website.
Click 2 of 5 · Upload
-
3
Let Ragety read them
Splitting, vectors and ranking happen on our side. You watch a progress bar. You do not select or operate a vector database.
Click 3 of 5 · Start processing
-
4
Ask a test question
In the playground you see the answer and the exact passages behind it, and you adjust prompt and model, before anyone else sees a single answer.
Click 4 of 5 · Playground
-
5
Publish
One line of code on your website, or open the internal knowledge portal for employees. Every answer shows the source it came from.
Click 5 of 5 · Publish
Everything above happens in the browser. Nothing is installed and no database is set up on your side. The five minutes start once your workspace is open, and in this early phase we open it together with you.
A pilot in five minutes, then every part of the company.
The same knowledge layer covers one department and every knowledge management task in the house. Each stage runs on the same permission model and the same record of requests, so growing means switching on a channel rather than starting a new project.
-
1
One project
What you get A playground and one published chat, with sources on every answer.What you operate Nothing. No database, no pipeline, no model hosting.Who is accountable fino data services operates the platform. You decide which content applies. -
2
Company-wide
What you get Projects per department, permissions per document, a record of every request.What you operate Still nothing. Content sync is triggered on demand or from your publishing pipeline.Who is accountable One contracting party for all projects, not one contract per tool. -
3
In your own product
What you get A REST interface and MCP for requests from your own application and from AI tools, with Slack to follow.What you operate Your product. The retrieval layer stays with us, including re-sync.Who is accountable Answers reaching your customers stay your editorial decision. Operation stays ours. -
4
From answering to doing
The idea An agent does not only answer, it acts. Ragety supplies the knowledge, the permissions and the record.What that needs The MCP connection from stage 3. The agent runs in your own tooling, with actions you approve in advance.Who is accountable Unchanged. You decide what an agent may do and where a human signs off.
Who Ragety is for.
For teams where the same questions keep coming up and the answer is sitting in a document.
-
Customer service
Your help centre answers questions itself and points to the article behind the answer. Agents get the passage instead of searching for it, and the team keeps the cases that genuinely need people.
-
Sales
Products, features, pricing and the roadmap in one internal knowledge scope. Reps prepare a call in minutes instead of asking product, and every answer comes with the document behind it.
-
Legal and compliance
Notice periods, clauses, policies. Questions are answered from the current version of the document, with the passage attached, so an answer can be checked instead of believed.
-
HR and people
Holiday rules, onboarding, company agreements. New joiners ask the handbook directly, and the team stops answering the same question every week.
-
Teams building knowledge management
Any department that maintains a body of knowledge and wants it used. Ragety turns documents that already exist into answers, without a migration project.
-
AI in customer communication
Teams putting AI in front of customers get a layer that only answers from approved content and shows where each answer came from.
Where the knowledge comes from and where the answers go.
Your existing systems remain the system of record. Ragety reads from them and puts the answer layer on top.
Sources Ragety reads from
| Source | What is taken in | Status |
|---|---|---|
| Documents | PDF, Word, PowerPoint, HTML, Markdown and plain text, individually or in bulk | available |
| Website and help centre | Content via the sitemap, re-synced after publication | available |
| Free text | Knowledge that is not in any document, entered directly | available |
| Spreadsheets | Excel and comparable table formats, including the values in the cells | planned |
| Knowledge systems | Confluence, Notion and comparable systems via connectors | planned |
| Your own systems | Connecting your own repositories through an interface | planned |
You decide when the sync runs. On demand, or triggered automatically from your own publishing pipeline as soon as content changes.
How answers reach people
| Channel | What it is for | Status |
|---|---|---|
| Chat on your website | One snippet of code on the page, appearance configurable per project | available |
| Internal knowledge portal | The same chat for employees, with its own knowledge scope and its own permissions | available |
| Playground | Test questions, review passages, set prompt and model | available |
| REST interface and MCP | Requests from your own application or from any AI tool, answers returned with their passages | available |
| Slack | Questions in a channel or by direct message, answers with the source | available |
The questions your IT security team will ask.
fino data services is certified to ISO 27001, builds software that more than 100,000 companies work with, and runs systems where mistakes touch money and duties of evidence. Ragety is built under the same rules. What that means in practice is set out here rather than in the small print.
Operated in the European Union
Your content can be processed and stored in the European Union. For cases with particular requirements, operation inside your own environment is foreseen.
Permissions down to a single document
You define which content belongs to a project and who may query it. A public and an internal assistant can therefore hold different knowledge.
A record of every request
Question, answer and the passages used are logged. You can trace at any time what was answered on what basis.
Separated workspaces
Every project is logically separated from the others. Your content is not used to train shared models.
You choose the language model
Models via AWS Bedrock or self-hosted models. Changing one does not change the integration in your applications.
Developed and operated to ISO 27001
Ragety is built inside the certified management system of fino data services, under its rules for access, change and continuity. The certificate is verifiable in the IAF register.
Transparency under the EU AI Act
The chat identifies itself as an AI system, every request is logged, and each answer stays tied to the passage it came from. Article 50 of the AI Act has applied since 2 August 2026, and these are the parts of it a platform can supply. Whether your own use of the chat falls under further duties stays your assessment. We support it with the record of requests and with a description of what the system does.
„We have spent more than ten years running systems where mistakes cost money. Ragety is built with the same attitude. Anyone who gives us their documents should know who processes them and who to call when something is wrong."
What your vendor assessment gets from us.
In larger companies, assessing the supplier often takes longer than deciding on the product. These documents are ready before you ask for them.
| What gets asked for | What you get | Status |
|---|---|---|
| Certified management system | ISO 27001, verifiable in the public register of the International Accreditation Forum | on file |
| Data processing agreement | Template under Article 28 GDPR, including technical and organisational measures | on file |
| Register of sub-processors | Complete list of the service providers involved, with location and role | on file |
| Place of processing | The region in which content is processed and stored | on file |
| Roles and permissions concept | How access is controlled down to a single document | on file |
| Logging and traceability | What is logged for every request and for how long | on file |
| Deletion concept | Retention periods for content and conversations, procedure at the end of the contract | on file |
| Handling of your content | Written commitment that your content is not used to train shared models | on file |
| Penetration test | Summary of the most recent test by an independent provider | on request |
| AI Act system description | What the system does, what it is intended for, where its limits are, and the transparency measures under Article 50, as input for your own assessment | on request |
If your organisation uses its own questionnaire, we fill it in. Just tell us who is responsible on your side.
Frequently asked questions
For a clearly bounded body of knowledge the five steps take about five minutes once your workspace is open. The platform is running and can be shown on real content. A few delivery channels are still in progress, Slack among them. We currently work with a limited number of companies whose use cases feed into development.
The name comes from the technology behind it: retrieval augmented generation, RAG for short. Your own content is searched before an answer is written. The first syllable is therefore pronounced like RAG, with a hard G, and not like the English word for anger.
RAG as a Service means a provider operates the retrieval pipeline for you: reading documents, splitting them, creating vectors, finding the right passage and calling the language model. You connect content and get answers, without running a vector database. Ragety is such a platform, operated by fino data services in Germany, with permissions per document and a record of every request.
A classic search finds documents that contain your words and leaves the reading to you. Ragety searches your content by meaning, also finds passages that use different terms, and writes an answer from them with a reference to the passage. You get the information rather than a list of hits.
Ragety answers only from the content you connected and discloses the passages behind every answer, so a check takes seconds instead of hours. Users rate answers, and the analytics show how often the chat was used and the ratio of good to poor ratings. An automated check against stored test questions is planned.
Ragety can be operated in the European Union, and your content is not used to train shared models. fino data services is certified to ISO 27001, verifiable in the public register of the International Accreditation Forum. We sign a data processing agreement under Article 28 GDPR including the technical and organisational measures, and disclose all sub-processors.
Yes. For each project you define which content it contains and who may query it, down to individual documents. Every project is logically separated from the others. A public chat on your website and an internal chat for employees can therefore run on the same platform and still hold different knowledge.
You replace nothing. Your wiki, intranet and ticket system stay the system of record, and Ragety adds the answer layer on top. Copilot is strong inside the Microsoft world. Ragety covers content outside it, references the passage behind each answer, keeps a searchable record and plugs into your own product. Many companies run both.
You choose the model. Ragety can call models via AWS Bedrock and equally use self-hosted ones, and switching does not change your integration. Article 50 of the EU AI Act has applied since 2 August 2026: the chat identifies itself as an AI system, and the record of every request supports disclosure duties. Your own legal assessment still applies.
Ragety is at an early stage and there is no public pricing yet. Scope is agreed in conversation, based on the volume of content, the number of requests and the operating model. As for the provider: fino data services is certified to ISO 27001 and has run systems for more than ten years where mistakes touch money and duties of evidence.
RAG as a Service, without running the pipeline yourself.
A retrieval layer someone else operates. What the service covers, and who is accountable for it once answers reach your customers.
RAGaaS, short for RAG as a Service, means a provider runs the complete retrieval augmented generation pipeline for you: reading your documents, splitting them into passages, creating vectors, storing them, finding the passage that answers a question and handing it to a language model. Ragety is such a platform, developed and operated by fino data services GmbH in Germany, with permissions down to a single document and a record of every request.
What the service covers
- Reading documents. PDF, Word, PowerPoint, HTML, Markdown and plain text, plus website content over the sitemap.
- Splitting content and creating vectors. Preconfigured, adjustable per project.
- Storing and searching the vectors. No database to run on your side.
- Finding the passage that answers the question. Search by meaning, so passages using different words are found too.
- Calling the language model. Over AWS Bedrock or self-hosted. Switching one does not change your integration.
- Keeping content current. Re-sync on demand, or triggered from your own publishing pipeline.
- Permissions, ratings and a record of every request. Part of the platform, not something you build around it.
- Transparency under the EU AI Act. The chat states that it is an AI system, and every request is recorded.
One operator for all of it
After go-live you run none of the parts above. fino data services operates the ingestion, the vectors, the retrieval, the model call and the re-sync, and remains the contracting party for every project rather than one contract per tool. That matters at the moment an answer reaches a customer. The operation is ours, and the editorial decision about which content applies stays yours. fino data services has run business-critical systems for over ten years, as a DATEV interface partner and a Peppol Service Provider.
Under European rules
For companies in Germany and the EU, three questions come before any feature does. Where is the content processed. Is there a data processing agreement under Article 28 GDPR with all sub-processors disclosed. Can the provider be checked rather than believed. fino data services is certified to ISO 27001 and verifiable in the public IAF register, Ragety can be operated in the European Union, and your content is not used to train shared models. Ragety is compliant with the EU AI Act: the chat discloses that it is an AI system, and every request is recorded.
Learn moreBring a question, take away an answer with evidence.
Tell us about a use case from your company. We set it up on a section of your content and show you the answers with their sources. No contract, and no effort on your side beyond providing the content. The number of places at this stage is limited.
- We set up your use case together with you rather than sending you to a signup page.
- Your requirements for connections and permissions feed into development.
- You get a named contact at fino data services rather than a help form.