How to Share Your Expertise as a Knowledge Base, and Be Paid When It’s Used

If you are a consultant, a lawyer, an accountant, a trainer or simply the person everyone asks, you have probably written the same explanations dozens of times. Those answers live in guides, templates, checklists and notes that only you can navigate. A knowledge base is a new way to monetize your expertise: you upload those documents to a library of expert knowledge bases, a language model answers questions from them with citations, and you are paid when people (through a subscription) or their AI agents (per request) use it. This guide covers how it works, what content works well, how to price access for agents, and how to launch without overpromising.

What is a paid knowledge base?

A paid knowledge base is a collection of your documents made queryable in natural language, with access that is paid for: through a monthly subscription for people chatting on the website, per request for AI agents and applications. Instead of buying a course or a book and searching through it, the user asks a precise question and gets a precise answer, with numbered references to the passages it comes from.

The technology behind it is retrieval-augmented generation, or RAG: the system searches your documents for relevant passages, then a language model writes an answer grounded in them. If the concept is new to you, start with What is RAG?. What matters for monetization is simple: your documents remain the source, and every answer is traceable to them.

Why a knowledge base instead of a course, an ebook or consulting hours?

Each format has its place. A knowledge base is not a replacement for your consulting work or your training programs; it is a different product, aimed at people who have a specific question right now and do not need, or cannot afford, a full engagement.

• A low barrier for users. A monthly subscription to a whole library, or a few cents per request for an agent, is an easy decision, much easier than buying a course.
• You reuse what already exists. Your templates and notes become a product without being rewritten as lessons.
• It works while you sleep. Questions are answered without you being involved in each one.
• It reaches AI agents. More and more work is done through assistants. A base available through an API or MCP can be consulted by those agents, which opens a channel courses and ebooks do not have.

What kind of expertise sells well as a knowledge base?

The best candidates share three traits: people have precise, recurring questions; the answers are written down somewhere; and getting them wrong has a cost. Some examples of domains where this pattern shows up:

• Regulation and compliance: a clear digest of a specific regulation for a specific industry, such as data protection for small online shops or safety rules for event organizers.
• HR and employment: practical guides to employment rules for a given sector or country, contract templates explained clause by clause.
• Tax and accounting: guides for freelancers, VAT scenarios, year-end checklists for small businesses.
• Real estate: rental rules, due diligence checklists, landlord and tenant obligations.
• Technical know-how: product documentation, integration guides, troubleshooting playbooks.
• Education: exam preparation material, structured notes for a certification.

How to build a knowledge base worth paying for

The mechanics of creating a base are covered step by step in how to build a knowledge base from your documents. For a paid base, the bar is higher, because buyers judge you on every answer.

1. Define the promise in one sentence. Who is it for, and what questions does it answer? This becomes your base description.
2. Select only authoritative content. Your best, most current documents. Remove drafts and anything you would not stand behind.
3. Fill the gaps. List the 30 questions your clients ask most and check that the answer to each is written somewhere in the corpus. If not, write a short document that answers it.
4. Write for retrieval. Explicit headings, defined acronyms, one topic per section. Our guide on preparing documents for AI goes into detail.
5. Date your content. Mention the date or version inside documents so readers know how current an answer is.
6. Test like a skeptical customer. Ask tricky, ambiguous and out-of-scope questions, and read the cited passages.

How to price access for AI agents

On Kopik, the price you set applies to requests from AI agents and applications, over the REST API and MCP. People chatting on the website use a subscription that covers the whole library, and you receive a fixed share of each question they ask your base. For the per-request price, reason from the buyer's point of view: what is a good answer worth to them, compared with the alternatives (searching for an hour, calling a professional, taking a risk)?

A few principles apply regardless of the topic. Start lower than you think, because a low price encourages people to try and builds usage. Raise the price once you see that people come back. And remember that AI agents may ask several questions in a row to complete a task, so a very high price per request can discourage automated use.

Share your expertise in the Kopik library

Kopik is the library of expert knowledge bases for AI, made in France. You create a base from your documents (PDF with a text layer, Word, and text formats such as .txt, .md, .csv or .html, up to 4 MB per file, or simply pasted text). Creating a base is free, and the platform handles extraction, indexing and answering with numbered citations.

You also choose the visibility: public (listed in the catalogue under a topic such as legal, HR and employment, tax and accounting, health or real estate), unlisted (accessible only via its link, useful for sharing with an existing audience), or private (only you can query it). A good habit is to start private: your own questions are free, so you can test the base thoroughly before opening it to others. Public and unlisted bases get their own page, which you can share on your website, in a newsletter or on social networks; it shows the names of your documents, never their full content.

Reaching AI agents, not just humans

This is the part that makes paid knowledge bases different from previous digital products. AI assistants and agents (Claude, Cursor, ChatGPT or an agent a developer built in-house) can connect to external tools through the Model Context Protocol (MCP). When your base is available as an MCP tool, an agent working on a task can consult it, get a sourced answer, and continue, with each request billed at the price you set.

On Kopik, buyers connect through a REST API or the MCP server using their own API key; a "passages" mode even returns the relevant excerpts without a written answer, for agents that prefer to reason on the raw text. The setup is explained in how to connect a knowledge base to Claude with MCP and in the developer documentation. Buyers can also set a maximum price per call (the maxPriceCents parameter): if a base costs more than that, the call is refused without charge. For an agent that queries your base in a loop, this is reassuring, since a later price change can never cost it more than it agreed to. As the creator, you do not have to build any of this: it comes with every base.

Use the questions to improve your base

Once the base is live, your dashboard shows the text of the questions people ask (not who asked them) and flags the ones that found nothing in your documents. That list is the most valuable feedback you will get: each unanswered question is either out of scope, which you can state in the description, or a gap you can fill with a short new document. The base grows with every upload, and the next user gets an answer.

Your earnings are shown in the same dashboard. They become payable 90 days after the question that generated them, a delay that covers payment disputes, and are paid out by bank transfer on request from your dashboard. Earnings from fraudulent questions or disputed payments can be withheld.

How to promote your knowledge base

A place in the library helps people discover you, but your existing audience is usually your best first market. They already trust you.

• Link it where questions arrive: your website's contact page, your email signature, your FAQ. "Have a quick question? Ask my knowledge base."
• Use it as a lead magnet: the two free questions each month on the website let prospects experience your expertise before hiring you.
• Write a clear description: who it is for, what it covers, what it does not cover, and when the content was last updated.
• Share sample questions: show three or four questions the base answers well, so visitors understand its value immediately.
• Keep it fresh: announce updates when rules change. A maintained base is worth more than an abandoned one.

Pitfalls to avoid

• Overpromising. A knowledge base provides information from your documents, not personalized professional advice, and an answer can be wrong. Say so in the description, especially on legal, tax or health topics. Our article on RAG for legal, HR and compliance covers the specific precautions.
• Letting content go stale. Outdated answers damage your reputation faster than no answers.
• A base that is too thin. A handful of pages rarely justifies payment. Make sure the core questions are covered in depth.
• A base that is too broad. Mixing unrelated topics dilutes retrieval and blurs your positioning.
• Pricing without testing. Watch how people use the base before optimizing the price.

Launch checklist

1. Write the one-sentence promise and the target audience.
2. Gather your best, current documents and check your rights to them.
3. List the top 30 client questions and make sure each is answered in the corpus.
4. Create the base as private, set its document language and topic, and upload the files.
5. Test with real, tricky and out-of-scope questions (they are free on your own base).
6. Set an initial per-request price for agents and switch the visibility to public or unlisted.
7. Write the description, with scope, limits and last update date.
8. Share the base page with your audience and link it from your site.
9. Review the questions asked, especially those that found nothing, and update the content regularly.