If you've ever typed "write me a high-protein grain bowl for 200 covers" into ChatGPT, you already know how good AI is at the first draft. You also know what happens next: you copy the result into a spreadsheet, guess at the nutritionals, re-key it into whatever system your team actually cooks from, and lose the thread the moment the recipe changes.
The draft was never the hard part. The hard part is everything that makes a recipe real: the cost per portion at your prices, the yield after trim and cook loss, the allergens that come from the vendor item you actually bought, the nutrition panel your client or regulator will ask for. That information lives in your operation. It should live in a system you control, and you should be able to take it anywhere.
That's the idea behind the Galley CRP platform's open API and MCP server. Your recipes, menus, ingredients, costs and nutrition data are yours. Ask questions about them in Claude. Build a report in ChatGPT. Wire them into your own tools. The data doesn't belong to the chatbot and it doesn't belong to us.
Food data has always been locked in the wrong places
Ask a culinary director where their recipe data is and you'll get a list: a shared drive of PDFs, a costing spreadsheet one person understands, the nutrition analysis from a consultant two menu cycles ago, a binder in each kitchen. Purple Carrot's product team described it plainly before they consolidated: "Our recipe data had to go many different places... it was a lot. People weren't able to see the bigger picture."
AI chatbots quietly added one more silo. The recipe you developed in a chat window in March is somewhere in a scroll of conversations. It has no cost, no allergen data, no version history and no link to what the kitchen produced. It's a sketch, not a record.
Culinary Resource Planning (CRP) exists to fix the first problem: one structured source of truth for ingredients, recipes, menus, production, inventory and procurement, where a change in one place updates everything downstream. An open API and an MCP server fix the second. They make that source of truth available to the AI tools your team already likes using.
What "open" means here
Two doors into the same data.
The open API is the developer door. Your engineering team, your BI analyst or your integration partner can read and write recipes, ingredients, menus, inventory and more. It's how operators push recipe data to point-of-sale, pull costs into finance, or feed menu data to partners like Leanpath for food-waste costing. If you can write a script, you can get your data out in the shape you need. The Galley API documentation is public.
The Galley MCP server is the everyday door. MCP (Model Context Protocol) is the open standard that lets AI assistants connect to outside systems. Connect Claude, ChatGPT or another MCP-capable assistant to your Galley account and you can ask questions in plain language:
"Which ingredients in the Galley Burger are low in stock right now?"
"Show me every recipe over 30% food cost on next week's cycle menu."
"Which menu items on Thursday contain tree nuts?"
"Increase all menu item volumes for tomorrow by 10%."
The assistant reads (and, if you allow it, edits) live data through the same API. No export, no CSV, no re-keying.
You decide how much access the AI gets. The MCP connection can be read-only, limited to specific kinds of edits, or given full write access. An intern's assistant can look but not touch. A culinary director's assistant can rebalance a production plan. You set the boundary.
Why this is a better deal than "AI inside the app"
Every software vendor now has an AI button. Most of them are a chatbot bolted onto one product that only knows that product. Galley Assist, the AI layer inside the CRP platform, does real work. It parses a recipe from a prompt, a file or a URL, imports nutrition labels and spec sheets from images, and builds menu plans from a catering export. We wrote more about these agentic integrations. But we don't think the only AI you'll ever want to use is ours.
You might prefer Claude for analysis and ChatGPT for drafting. Your finance team might already have an AI workflow for variance reporting. Your IDE's coding assistant might be the fastest way to build the integration you've been putting off. An open API and an MCP server mean all of those tools work with your real food data, and you keep the freedom to switch.
That's the "from and to" of this whole category. From: data trapped in binders, spreadsheets and chat logs. To: one connected culinary operation whose data you can take anywhere.
What operators actually do with it
A few patterns we see once the data is open:
Costed menu reviews in minutes. A director connects Claude to their account and asks for every item on the spring menu sorted by margin, then asks what happens to margin if the salmon price rises 12%. Both answers come from live recipe costs and vendor items.
Allergen and nutrition questions answered by anyone. A dietitian lets the front-of-house team ask a read-only assistant "what on today's menu is gluten-free and under 600 calories?" The answer traces back to the actual recipe build, not a guess.
AI recipe development that ends in the system of record. A chef drafts in ChatGPT, imports the draft into the platform with Galley Assist, and gets back a recipe with sub-recipes, yields, cost and an FDA-style nutrition panel.
Integrations built faster. A developer adds the Galley MCP server to their IDE so the coding assistant can explore the API and write the integration with them.
The principle
You did the work of building your recipes, your costs and your nutrition data. A platform that holds that data should make it easier to use everywhere, not harder to leave. Open API. Open protocol. Your data, your choice of tools.
See how the Galley MCP server connects your kitchen data to Claude or ChatGPT. Book a 20-minute walkthrough.
Frequently asked questions
Does Galley have an open API? Yes. The Galley CRP platform exposes an API for reading and writing recipes, ingredients, menus, inventory and more. The documentation is public: Using the Galley API.
What is the Galley MCP server? An MCP (Model Context Protocol) server that lets AI assistants such as Claude or ChatGPT read and, with permission, edit your Galley data through natural-language prompts.
Can I limit what an AI assistant can change? Yes. Access can be set to read-only, limited edits or full edits.
