Connect Your Own AI to Net Net with the MCP Server

In brief: You point your own AI client at Net Net's MCP server and it reads your live workspace under your own permissions.


Applies to: Anyone on an active trial or a paid plan. MCP is not available on the free tier.


See also: Setting Up the MCP Server in Claude, ChatGPT, and More covers the connection steps, What to Ask Your AI About Your Net Net Workspace covers what to do once it is live, and AI Access and Your Plan: Why MCP Needs a Paid Plan covers the plan requirement.


Net Net does not ask you to learn a new chatbot. It opens your workspace to the AI assistant you already pay for and already trust, over an open standard called MCP. You point Claude, ChatGPT, Gemini, Perplexity, or Grok at one address, sign in with your Net Net credentials, and then ask questions about your own Jobs, your own effort, and your own contacts in plain language. Net Net supplies the data and the domain model. Your AI supplies the reasoning. It requires an active trial or a paid plan.


The address you connect to is https://mcp.netnet.app/mcp.


This article explains the idea. When you are ready to connect something, the step by step is in Setting Up the MCP Server in Claude, ChatGPT, and More. For a step by step walkthrough of connecting Claude specifically, watch the video near the top of that article.


What MCP actually is


MCP stands for Model Context Protocol. Behind the name it is a doorway with a lock on it.


Most AI assistants are excellent at reasoning and completely blind to your business. Yours can explain what drift is. It cannot tell you which of your Jobs is drifting, because it has never seen your data. Pasting a report into a chat window closes that gap for about five minutes, and then the report is stale and you are pasting again.


MCP is the standing connection that removes the pasting. Your AI asks Net Net a question, Net Net answers with live data from your workspace, and your AI reasons over the answer. Nothing is exported to a third party. Nothing is synced overnight into a copy that quietly goes out of date. What your AI sees is what is in your workspace at the moment you ask.


The lock matters as much as the doorway. You authorize the connection once, using your own Net Net login, and you can remove it from your AI tool whenever you want.


What bring your own AI means


Most software that adds AI adds its own. You get a chat panel, a new interface to learn, a new set of quirks to work around, and usually a new line on the invoice.


Net Net went the other way. You already have an assistant you have trained yourself on. You know how it phrases things, where it is strong, where it needs a second prompt. That relationship took months to build and it is worth keeping.


So Net Net does not compete with it. Net Net hands it your workspace.


You choose the assistant. Claude, ChatGPT, Gemini, Perplexity, Grok, or whatever replaces them next year. The protocol is open, so any assistant that speaks MCP can connect.


You keep your existing plan. You are already paying your AI provider. Connecting to Net Net does not add a second AI bill, because Net Net is not running a model on your behalf. Every token you spend is spent with your provider, on your terms.


You keep your other context. Your assistant does not stop knowing everything else it knows the moment it connects to Net Net. It can hold your workspace data alongside the proposal you were drafting, the client email thread you pasted in this morning, and whatever else lives in that conversation.


The reasoning behind that choice is told at more length in the AI, Honestly section of the Net Net workflow management software page.


Why this beats a built in assistant


A built in assistant is frozen at the level of whatever the vendor shipped. When the models get better, and they get better constantly, you wait for a release.


Bring your own AI inverts that. The intelligence layer upgrades on your provider's schedule, not on Net Net's. When your assistant gets a better model next month, your Net Net questions get better answers next month, and Net Net ships nothing at all.


There is a division of labor underneath this that is worth naming, because it is the whole design:


Net Net provides the data and the shape of it. Not a raw table dump. Your AI receives the structure that makes the numbers mean something, so it understands that a Job holds Deliverables, that Deliverables hold pools of planned effort by service type, and that logged time is measured against those pools rather than against a vague total.


Your AI provides the intelligence. The reading, the pattern finding, the summarizing, the drafting, the awkward follow up question you would not have thought to ask.


Neither half is very interesting alone. A workflow tool with no reasoning gives you dashboards you have to interpret yourself. A brilliant assistant with no access to your data gives you generic advice about agencies in general. Together they answer the question you actually have, which is how your work is going, right now, in your shop.


Net Net bears no model cost on MCP usage. You pay your own AI provider directly, at whatever rate your plan with them already sets.


It acts as you, never above you


This is the part worth reading twice.


When you connect your AI over MCP, you authorize it with your own Net Net login. From that point it is operating as you. It carries the permissions of your role, and Net Net enforces those permissions on the server for every request, exactly as it does when you click through Net Net yourself.


A Member connected over MCP can see and do what a Member can see and do. Not more. If a report is closed to your role in the product, it is closed to your AI. There is no separate AI account with elevated reach, and there is no back door around the permission layer.


The honest flip side deserves saying out loud: you also cannot give your AI less access than you have. The connection inherits your role whole. It is not currently possible to hand your assistant a read only slice of a workspace you can write to. If your role can delete a Job, an assistant connected as you can delete a Job.


So treat the connection the way you would treat handing someone your login. That is genuinely what it is. Net Net's role differences are explained in User Roles Explained: Owner, Admin, Lead, and Member.


Your AI can create, change, and delete real records in your workspace, up to the limit of your own role. Read what it proposes before you tell it to go ahead, particularly for anything involving deletion or bulk changes.


What you need before you start


An active trial or a paid plan. MCP is not available on the free tier. It is part of the 60 day trial and part of every paid plan. The full detail is in AI Access and Your Plan: Why MCP Needs a Paid Plan.


An AI assistant that supports custom remote connectors. Most of the major ones now do, though the required tier on their side varies quite a lot. That is covered provider by provider in Setting Up the MCP Server in Claude, ChatGPT, and More.


Your normal Net Net login. There is no API key to generate, copy, or store. You sign in through your AI tool with the same credentials you use every day, and the authorization happens between the two systems.


A workspace with something in it. An empty workspace gives your AI nothing to reason about. Plan a Job first, log a little time, and the questions get far more interesting. Getting Started With Net Net: How It All Fits Together covers that ground.


Frequently Asked Questions


Do I need a separate AI subscription to use this?


You need whatever plan your chosen assistant requires for custom connectors, and that is the only AI cost involved. Net Net does not charge for MCP access beyond requiring a trial or paid Net Net plan, and Net Net does not resell you tokens. You pay your AI provider directly, which also means your usage is governed by their limits rather than ours.


Is my workspace data sent somewhere and stored?


Your AI reads your workspace live, at the moment you ask, through the connection you authorized. Net Net does not push a copy of your data to your AI provider in advance. What happens to a conversation after your assistant answers is governed by your AI provider's own data policy, so if that matters to you, read theirs.


Can I connect more than one assistant?


Yes. Each assistant authorizes separately and holds its own connection. You can run Claude and ChatGPT against the same workspace if you want to compare how they answer, and removing one does not affect the other.


What can my AI actually do once it is connected?


It can read your Jobs, Deliverables, tasks, logged time, and contacts, and it can create and change records that your role allows. Concrete examples, grouped by what you are trying to find out, are in What to Ask Your AI About Your Net Net Workspace.


Is there a chatbot built into Net Net?


No. There is no chat panel or assistant inside the product, and connecting your own over MCP is the whole AI story today. That is a deliberate choice rather than a gap waiting to be filled: the assistant you already use improves on your provider's schedule and already holds the rest of your working context, which a built in panel never would.


What happens to the connection if my plan lapses?


The connection stays configured in your AI tool, but Net Net stops answering its requests. MCP access is checked against your workspace status every time your AI asks a question, so a workspace that has returned to the free tier will have those requests declined. Restore an active plan and the same connection starts answering again. See AI Access and Your Plan: Why MCP Needs a Paid Plan.


Related articles


If you have not connected anything yet, Setting Up the MCP Server in Claude, ChatGPT, and More walks through where the setting lives in each provider, and now carries a video walkthrough of the Claude path. Once the connection is live, What to Ask Your AI About Your Net Net Workspace is the list of questions worth starting with. AI Access and Your Plan: Why MCP Needs a Paid Plan covers what happens to the connection when a trial ends or a payment fails. And if Job, Deliverable, and effort do not yet mean anything precise to you, The Net Net Vocabulary: Every Term in the Product is the fastest way to fix that, because your questions land better when you use the same words the data does.


Updated on: 08/04/2026

Was this article helpful?

Share your feedback

Cancel

Thank you!