Why we built the Secfi MCP
13 min
Paris born, based in Amsterdam. I manage product and tech at Secfi. Passionate about everything design, tech and finance.
0 result
0 result
Paris born, based in Amsterdam. I manage product and tech at Secfi. Passionate about everything design, tech and finance.
Behind the build: why having Secfi’s calculators in existing LLMs like ChatGPT can help more people get better information for their taxes and equity decisions.
Many of us are already using LLMs like ChatGPT daily, whether it's for comparing phones or figuring out how to handle a tricky conversation with a coworker. And increasingly, we’re using them to make sense of financial decisions too, like understanding our stock options and estimating taxes.
But even our favorite general LLMs have their limits. They’re prediction engines, not calculation engines. They’ll fill in the blanks somewhere along the line, then sometimes give you an entirely different answer if you change the phrasing slightly. And when the numbers can have a major impact on your finances, you don’t want “probably right”...you want mathematical certainty.
When we built our AI equity assistant Maeve, we knew we could use our calculators to replace overly confident AI replies that might not reflect your reality.
Now, we’re taking the next step.
We built the Secfi MCP so you can connect Secfi’s equity data and financial calculation tools directly to AI assistants. Instead of moving between a general AI tool and a separate equity platform, you can ask questions in the AI interface you already use and let it pull in relevant information and calculations from Secfi.
We’re pretty excited about it, so here’s a bit more about our free MCP.
Note: Secfi helps startups, executives, and employees understand and act on their equity. Try our free AI equity assistant Maeve today, or log in to ChatGPT to try the new MCP!
MCP stands for Model Context Protocol. It’s an open standard that lets generalist AIs communicate with external, specialized services. In Secfi’s case, we built a set of tools that allow ChatGPT to request information and calculations from our platform.
When we built Maeve, we knew the next opportunity was to make its calculators and logic accessible inside the most widely available AIs.
We get it, a lot of us don't want to leave our favorite LLM ecosystems. It becomes part of your comfort zone, versus logging into a different system every time you have a question about your equity.
That means we want our equity tools available wherever our customers choose to work, whether that’s ChatGPT or another AI platform in the future. With the Secfi MCP connected, you can ask your preferred AI to pull in your Secfi data and run equity calculations.
The personalization that comes from your preferred LLM can give context to every equity decision.
If ChatGPT already knows that you work in UX at a startup but dream of starting your own digital art cafe, while you’re also trying to pay off student loans and send money for your parent’s healthcare…it’s going to pick up on things that might matter in your equity decisions.
Even if you give all your grant information to Maeve and a bit of backstory, it won’t have the years of information you’ve shared with your AI tool of choice. All of that background can help when making decisions about how much cash you’re comfortable tying up in an exercise, or how much risk you want to take.
General-purpose AI tools are very good at explaining concepts and holding a conversation. But equity planning asks them to do something they weren’t primarily built for: produce consistent financial calculations from complicated, incomplete information.
A few things can go wrong with standard LLMs:
The Secfi MCP gives ChatGPT access to tools built specifically for these problems, including Secfi’s tax and equity calculation engines, your connected equity data, and private-market insights. In that sense, you get the best of both worlds: the conversational experience and broader context of your preferred AI assistant, with Secfi’s specialized equity data and calculation tools behind it.
That’s the same basic principle behind Maeve, our dedicated AI assistant. The difference is that the MCP brings those Secfi capabilities into an AI environment you may already use every day.
That being said, we still recommend using Maeve to get the most out of our specialized agent. With our dedicated app, you not only get access to Maeve, but also purpose-built tools and visualizations that can help you to better interpret and zoom in on your equity position.
So, what actually happens when you connect Secfi to ChatGPT?
It’s similar to other connectors you might have tried in your LLM. For example, in ChatGPT, type “Secfi” into connectors. Once you’ve added it, you can call on Secfi’s calculators in your regular chats.
The easiest way to think about it is:
You ask the question → ChatGPT contacts Secfi → Secfi returns the relevant data or calculation → your AI assistant uses that result in its answer.
For example, imagine you ask:
“What would my stock options be worth if my company exits at $20, $50, or $80 per share?”
Instead of asking the language model to estimate everything itself, the AI can call Secfi’s exit calculation tools for each scenario, get the results back, and then help you compare the outcomes. The more information you’ve connected, like grant details and tax profile, the better the answers.
The MCP gives the AI a way to pull in the relevant Secfi data and calculations as the conversation develops, instead of making you manually copy every number into a prompt each time.
Note that you’ll need to create a free Secfi account to use the MCP. Get started if you don’t already have one. When you’ve logged in, the MCP works within your AI for 30 days before needing you to authorize it again.
Once Secfi is connected, you can start asking the kinds of equity questions that usually require jumping between calculators, spreadsheets, company documents, and general AI tools.
Here are a few examples.
Alternative minimum tax (AMT) is a federal tax system that can apply when you exercise incentive stock options (ISOs), even if you haven’t sold the shares yet or received any cash. AMT can be especially difficult to estimate because the outcome depends on several inputs at once, and strategies to reduce it often involve comparing different exercise amounts and timing scenarios.
For many startup employees with ISOs, one of the biggest questions is whether exercising will trigger AMT, and how large that tax bill could be. In some cases, that can lead to an unexpectedly high tax bill if you haven’t planned for it.
Use Secfi’s tax calculation tools and your actual equity information to ask:
Private-company shares don’t have a single live market price. Information from secondary transactions, funding rounds, and other private-market activity can also be difficult to find and consolidate. Secfi’s MCP can bring in private-market insights and data from sources like PitchBook and Caplight alongside your own equity information.
You could ask:
Your stock options rarely exist in isolation from the rest of your finances.
If your preferred AI already knows about your budget, home-buying plans, student loans, or other goals, Secfi’s MCP can bring your equity data and calculations into that wider conversation too. You can ask:
Leaving a startup can make your equity suddenly time-sensitive. Depending on your grant and company plan, vested options may need to be exercised within a post-termination exercise window, which is often 90 days but can vary.
Secfi’s data and calculation tools can help you understand your vested options, potential deadlines, exercise costs, and possible tax implications. You might ask:
Using LLMs for financial planning will likely be as common as our go-to spreadsheets for many in the near future. But instead of relying on predictive text models, bringing calculation engines like those powering Secfi’s MCP can give users more confidence in their planning.
We think this model could eventually go much further than answering stock questions.
Building a financial plan can take hours. Then life changes, and parts of that plan need to be revisited. We’re working towards structured, multi-year financial plans with reliable calculations underneath them.
The underlying financial models can recalculate the numbers, while the AI helps you understand what changed and continue asking follow-up questions.
There’s also room for collaboration on financial plans in the future. For example, you might share a link with your financial plan or outputs to a partner or financial planner, so you can each work through scenarios for a major household financial decision.
Many of us already use AI to research, plan, and make sense of complicated decisions.
Instead of visiting a different website or app for every financial question, our MCP helps you stay inside of your preferred AI assistant and bring calculations into that environment when you need it.
And we think financial planning will increasingly work this way: specialized financial infrastructure underneath, with the interface chosen by the user.
Try our MCP today by logging into ChatGPT and adding it as a plugin or connector, or sign up for Maeve if you haven’t tried it yet.
The Secfi MCP server is an integration that lets AI assistants like ChatGPT access your Secfi equity data and equity planning tools, run Secfi calculations, and answer equity compensation questions without making you switch to another app.
Generalist AIs, like ChatGPT or Claude, are excellent at holding a conversation and helping you brainstorm about your needs and preferences. However, the moment you want to turn those preferences into hard numbers and calculations, they fall short. They are not engineered to provide consistent calculations and analyses, but rather to predict what the likely “next sentence” is.
The next sentence will probably sound realistic, and in some cases it may even be roughly right. But when making major financial decisions, is roughly right good enough?
Yes. You need to log in to your Secfi account before ChatGPT can use your personal equity data or run calculations based on your grants and tax profile. That can include information about stock options, RSUs, and other private-company equity you’ve added to Secfi. After you’ve connected to the MCP, you only need to log in again every 30 days.
Once connected, your AI assistant can access tools for grant details, portfolio summaries, tax information, exercise cost and AMT calculations, and exit scenario modeling.
You can also ask broader questions about equity compensation, such as how an IPO, secondary sale, or other potential liquidity event could affect your shares. For employees at late-stage private companies, that can help connect company developments with their own equity scenarios.
Install the MCP from the connector store, then ask your AI assistant to “Connect my Secfi account”. A browser window will open so you can log in, authorize the connection, and return to the chat. Secfi says the session stays active for 30 days before you need to reconnect.
The Secfi MCP can help you explore equity scenarios and understand your data, but its responses aren’t financial advice or a replacement for personalized professional guidance. We recommend that you talk to a tax professional and a financial advisor when it comes time to make the final decision.
If you need broader financial planning or wealth management support, Secfi Wealth offers wealth management services that can account for private-company equity alongside the rest of your financial life.
The tool shown here uses artificial intelligence and is for illustrative purposes only and not necessarily indicative of future results and there is no guarantee that similar results can be achieved. The information provided by the tool is not professional advice and is not intended by Secfi, Inc., its affiliates, and Secfi representatives, to be deemed as investment, legal, tax or other professional advice or recommendations of any kind, or to form the basis of any decision to do or to refrain from doing anything. Secfi does not review the accuracy or completeness of the information provided to us within the tool.