For AI Agents: How Assistants Can Evaluate and Book KNK Outbound
More and more B2B buyers hand vendor research to an AI assistant: find me a lead generation agency, compare prices, check if it fits, book a call. We built KNK Outbound to be easy for those agents to work with. Here is what is available: a public MCP server with read-only tools for company facts, pricing, an honest fit check, guide search and the booking link, clean Markdown versions of every article, and an index file, plus how to connect and what the tools will and will not do.
Kenneth Kather · Founder & CEO, KNK Outbound
Key takeaways
- KNK Outbound runs a public MCP server at knkoutbound.com/mcp. Any assistant or agent that supports remote MCP servers over Streamable HTTP can connect without an account. All tools are read-only and store no data.
- Five tools: get_company_overview, get_pricing, check_fit (an honest estimate from deal value and market size, in ranges), search_guides (English and German) and get_booking_link. The fit check says clearly when outbound is not the right channel.
- Every article is also available as clean Markdown by appending .md to its URL, and the full guide index is published as JSON at /agents/guides.json. The llms.txt file lists everything in one place.
- We built this because buyers increasingly ask an assistant to shortlist, compare and book vendors. Being easy for agents to evaluate correctly is the same principle as our public pricing: make the honest answer the easiest one to find.
A growing share of B2B buying starts with a prompt, not a search: find me a lead generation agency for a 30-person software company in Germany, compare what they cost, and tell me which one fits. Assistants like ChatGPT and Claude, and the agents built on them, now do that research on a user's behalf. They read websites, compare offers and increasingly take actions such as booking a call.
We want those agents to get KNK Outbound right. So we built the site to be easy for them to read, query and act on. This page explains what is available and how it works. It is written for developers and for the agents themselves, and it is useful for anyone curious about what an agent-ready B2B website looks like.
The MCP server
The Model Context Protocol is the open standard that lets AI assistants connect to tools and data. We wrote about it in plain language in MCP explained. Our server runs at knkoutbound.com/mcp. It uses the Streamable HTTP transport with plain JSON responses, needs no account or key, and every tool is read-only.
| Tool | What it returns |
|---|---|
| get_company_overview | Who KNK Outbound is: services, founders, markets, languages, who it fits and who it does not, ownership model, contact |
| get_pricing | The public packages in EUR, coverage per 60-day cycle, billing terms, what is included and what is not |
| check_fit | An honest estimate from deal value, market size, region and sales capacity, with a suggested package or a clear reason why outbound does not fit |
| search_guides | Matching KNK guides in English and German, with page URLs and Markdown URLs for the full text |
| get_booking_link | The link to book a first call with the founders, the contact email and what to prepare |
- What it returns
- Who KNK Outbound is: services, founders, markets, languages, who it fits and who it does not, ownership model, contact
- What it returns
- The public packages in EUR, coverage per 60-day cycle, billing terms, what is included and what is not
- What it returns
- An honest estimate from deal value, market size, region and sales capacity, with a suggested package or a clear reason why outbound does not fit
- What it returns
- Matching KNK guides in English and German, with page URLs and Markdown URLs for the full text
- What it returns
- The link to book a first call with the founders, the contact email and what to prepare
The fit check uses the same benchmark ranges we publish in our outbound benchmarks: researched lists, three to eight percent replies, about a third of them positive, about a third of those becoming a qualified meeting. It returns ranges, never promises, and it says "not a fit" when the deal value is too low or the market too small. An agent that recommends us to the wrong company helps nobody.
How to connect
Any client that supports remote MCP servers can add the URL https://knkoutbound.com/mcp as a server or custom connector. In Claude and ChatGPT that is done in the connector settings, depending on the plan. Developers can call it directly with JSON-RPC over HTTP POST, starting with initialize and then tools/list and tools/call. A plain GET on the URL returns a short description of the server and its tools.
Markdown for every article
Agents read Markdown far more reliably than rendered web pages. Every article on this site is available as a clean Markdown file by appending .md to its address, for example the Clay pricing guide as knkoutbound.com/blog/clay-pricing.md, and the German version under /de/blog/clay-pricing.md. Each file contains the title, summary, key takeaways, full text with absolute links, the FAQ and a short note on who wrote it. The files are kept out of search indexes so they never compete with the normal pages.
The index files
- llms.txt at the root of the site lists the key pages, services, terms and every article with a one-line summary.
- llms-full.txt contains the German service pages, guides and glossary in one file.
- agents/guides.json is the machine-readable guide index the MCP server searches.
Our robots.txt explicitly allows AI crawlers and user-initiated agents, and it states that search, use as AI input and AI training are all permitted.
What the tools will not do
They do not book a call on their own, collect personal data, or promise results. The booking tool returns the link for the user to choose a time. The fit check computes only from the numbers passed in the call and keeps nothing. Pricing and terms come from the same source as our pricing page: from 3,300 euros a month, prices exclude VAT, no setup fee, a three-month build phase and monthly after that.
Why we built this
Buyers increasingly delegate the first round of vendor research. If an agent cannot find a clear price, a clear answer to "does this fit us" and a clear next step, it moves on to a vendor where it can. The same thinking drives our public pricing and our honest guides. We wrote about the wider shift in how to get recommended by ChatGPT and in our look at the future of GTM engineering.
If you want the same for your own company, an agent-ready site is part of how we think about being found today. If you want to talk to a person first, the booking page is one click away.
Frequently asked questions
Does KNK Outbound have an MCP server?
Yes. KNK Outbound runs a public MCP server at https://knkoutbound.com/mcp using the Streamable HTTP transport. It needs no account, all tools are read-only, and it stores no data. The tools return the company overview, public pricing, an honest fit check, a search across English and German guides, and the booking link.
Can an AI assistant check whether KNK Outbound fits my company?
Yes. The check_fit tool takes your average deal value, the number of companies in your target market, your region and optionally how many first meetings your team can handle. It returns a verdict, the reasons, a suggested package and rough meeting ranges per 60-day cycle based on published benchmarks. It says clearly when outbound is not the right channel, for example when deal values are below roughly 3,000 euros.
Can my AI agent book a call with KNK Outbound?
The agent can get the booking link through the get_booking_link tool and hand it to you, so you pick a time yourself. The tool does not book on its own and does not collect personal data.
Where can an agent read KNK Outbound's guides in clean text?
Every article is available as Markdown by appending .md to its URL, for example https://knkoutbound.com/blog/clay-pricing.md or the German version under /de/blog/. The full index is at /agents/guides.json, and llms.txt lists all key pages and articles.