Built to edit, with your AI.
Every Hemrock model is built so Claude, ChatGPT, and other AI can read and edit it without breaking the math. And the calculation engine behind the models is callable directly, over an API and an MCP server, so your AI runs the real numbers instead of doing finance in its head.
One engine, every surface
The spreadsheet, the in-browser tool, the API, and the MCP server all run the same calculation core. A cap table an agent computes over the API matches the one in your spreadsheet, because it is the same engine underneath, not a second implementation that drifts.
That is what makes AI safe to use here. The proven structure gives every edit a place to show up, and the engine makes the math that has to be exact exact, so you verify the output instead of trusting it.
Two kinds of tools
Context, for the judgment
Primers, task prompts, sanity checks, and concept guides that teach your AI how a model is structured and how a calculation actually works. This is what makes an edit land in the right cell and a review catch the right error.
Compute, for the math
The cap table, exit waterfall, and fund economics engines, callable directly. An LLM doing this math in its head miscounts an option pool or converts a SAFE on the wrong basis. Handing it to the engine removes that failure mode.
Ways to connect
Prompt guide
Any AIA copy-paste context primer and ready-made prompts for setup, customizing, building scenarios, and verifying. Paste it into Claude or ChatGPT before you touch a model.
Claude Skill
FreeA drop-in Claude Agent Skill that loads a model's sheet map, prompts, and sanity checks into any Claude Project. No connector to set up, clone it from GitHub.
MCP server
Claude, ChatGPT, CursorA live connector for any MCP client. It serves the modeling context and concept guides, and runs the calculation engines directly from a conversation.
Developer API
RESTA bearer-authenticated REST API. Run cap tables, exit waterfalls, and fund economics from your own code, scripts, or agents. When an agent hits the paywall it gets a checkout link it can act on.
What your AI can compute
Three engines run over the API and MCP today. Fund Economics is free. Cap Table and Exit Waterfall come with the Cap Table & Exit Waterfall product, over the API and MCP both.
Cap Table
Fold a list of events into per-round snapshots and final ownership.
Exit Waterfall
Distribute an exit across the cap stack: preferences, participation, conversions, options, warrants.
Fund Economics
Forecast a fund: fees, carry, deployment, and gross and net returns.
Works with every model
The context tools cover every template. Pull the primer, prompts, and checks for the one you're editing.
Why start from a structure
Most people use AI to build financial models from scratch. I think that is backwards. AI is good at editing structure that already works, and bad at generating correct structure from nothing.
So start from a model 55,000 people have used, let AI do the line-level edits, and call the engine for the numbers that have to be right. That is the whole approach, and these are the tools that make it work.
