Commercial systems and the seams
Positioning, marketing, sales and customer success built as one machine, with an owner for every handoff between them.
Our lead numbers are up but pipeline is flat. Where is the problem?
MCP server
Nemanja's Brain
Talk to my brain. Everything I have written about B2B commercial systems, packed into an MCP server. Connect it to your AI assistant and ask it the questions you would ask me.
WHAT THIS IS
An MCP server, in plain words
MCP (Model Context Protocol) is the standard way AI assistants like Claude, ChatGPT and Cursor plug into outside sources. This server is one of those sources. Once you connect it, your assistant can search my writing on its own while it answers you, and it shows you which essay or post each point came from. You keep using the assistant you already have. Nothing to install on your side except the connection.
WHY IT EXISTS
Nine years of writing, one question away
I have been writing about positioning, marketing, sales and customer success since 2017: long essays on Substack and LinkedIn, and hundreds of shorter LinkedIn posts. Nobody scrolls through all of that, and I don't expect them to. With this server you don't have to. Ask about the problem you have today, and the assistant pulls the passages where I already worked through something similar.
WHAT IS INSIDE
466 pieces, 2017 to 2026
56
long-form essays (Substack and LinkedIn articles)
410
LinkedIn posts
812
searchable passages
| 2026 | 138 |
|---|---|
| 2025 | 65 |
| 2024 | 74 |
| 2023 | 49 |
| 2022 | 12 |
| 2021 | 29 |
| 2020 | 27 |
| 2019 | 3 |
| 2018 | 9 |
| 2017 | 4 |
Last updated: October 2026.
TOPICS
What you can ask about
Demand generation and attribution
Create pull instead of chasing leads, pass only real opportunities to sales, and measure with self-reported attribution.
How do I prove LinkedIn brings deals when our attribution tool says direct?
Pricing and offer economics
Know what a delivered day costs before positioning, price to the client's market, publish your prices.
We think our margin is 30%. How do I check whether it really is?
LinkedIn as a revenue channel
Rules on hooks, formats, language and founder-led posting, scored on the buyers they bring in.
What kind of LinkedIn posts actually work for a B2B founder right now?
Positioning and focus
Positioning is the root everything else inherits from. Strategy is mostly deciding what to stop.
Should we niche down or keep talking to a broader market?
PRINCIPLES
Five ideas that show up everywhere
- Look for the broken system in the seams between functions. When a metric goes red, the cause usually sits in a gap between teams.
- Kill the MQL. Use one shared definition of an opportunity, judged by whoever had the conversation, and give the handoff one owner.
- Build the economics sheet before positioning: know what a delivered day costs and which offer actually makes money.
- Price against the client's market, publish your prices, and never cut price in response to silence.
- Grade LinkedIn on what buyers write in a free-text "how did you hear about us" field.
WHAT AN ANSWER LOOKS LIKE
Example
How do I know if LinkedIn actually drives revenue?
- Add a free-text "how did you hear about us" question to your inbound form and read it every week. Source
- Make the field mandatory, and add "why would you like to work with us?" to see how much buyers already know about you. Source
- Read direct traffic, organic search and word of mouth as possible delayed echoes of LinkedIn. Pixels and last click miss most B2B touches. Source
- Self-reported attribution measures demand creation. You still need software to measure demand closing. Source
A shortened example. Your assistant writes its own answer from the passages it finds, so wording will differ.
HOW IT WORKS
Two tools, read-only
search_brain
Your assistant sends a question in plain words. The server ranks passages by keyword relevance and returns up to 6 of them (10 at most), never more than 2 from the same piece, each with a link to the original.
list_sources
Returns the full list of essays and post collections, with links, so the assistant knows what it can draw on.
The server only reads. It cannot change anything in your files or accounts, and it does not write answers itself: your assistant does that from the passages it gets. Because the ranking is by keywords, a question phrased in different words can bring different passages, and a good assistant will search more than once.
Server URL
https://mcp.nemanjazivkovic.co/nemanja-zivkovic/mcpHow to connect
Claude
Web or desktop
1. Open Settings, then Connectors. 2. Click Add custom connector. 3. Paste the URL above and give it a name, for example Nemanja's Brain. 4. Click Connect. 5. Enter your email, then the 6-digit code that arrives in your inbox. 6. In a new chat, ask a question and the assistant will use the brain when it fits. If it doesn't, tell it to use Nemanja's Brain.
Claude Code
In your terminal
Run this in your terminal, then run /mcp inside Claude Code and choose nemanja-brain to sign in with your email.
claude mcp add --transport http nemanja-brain https://mcp.nemanjazivkovic.co/nemanja-zivkovic/mcp
Other MCP clients
ChatGPT, Cursor and others
Add the same URL as a remote MCP server (in ChatGPT this sits under developer mode). The client opens a sign-in page, you enter your email and the code. Where custom connectors are available depends on your plan in each app.
PRIVACY
What I see
Sign-in is by email: you get a one-time code, no password. When you connect, the server stores your email address, and the IP address and browser you signed up from. When your assistant searches, the server stores the search text it sent (up to 500 characters) and which app sent it. If your assistant suggests a call with me and fetches my booking link, the server also stores the one-line topic it sends with that request. Your conversation stays with your assistant. The server only gets the search it writes. I use this to see what people ask and what the brain is missing. Only I can see it. If you want your account and history deleted, email nemanja@funkymarketing.net and I will remove it.
GOOD TO KNOW
Limits
- It knows what I have published. It doesn't know your company, so give it context in your question.
- Older pieces reflect what I thought then. The dates are in the sources, so check them.
- It is free to use.
- For a decision that matters, the brain is a starting point. A conversation is better.
Try asking
Our lead numbers are up but pipeline is flat. Where is the problem?
How do I prove LinkedIn brings deals when our attribution tool says direct?
Who should own the handoff between marketing and sales?