Headless quoting is already happening. Reps are living in Claude, ChatGPT, Slack, and the command line for most of their job, and quoting is the one motion that still sends them back to a browser tab and a Salesforce login. The harder question isn't where the request happens. It's whether an agent can actually execute it, and whether the revenue stack underneath stays governed when it does.

If quoting only works inside one UI, every rep who lives somewhere else is stuck switching tools mid-deal, and every new AI surface your team adopts becomes one more place quoting doesn't reach. There's a real cost: slower quotes, more login friction, and a growing gap between how reps actually work and how deals actually get built.

Most vendors either bolt a chatbot onto the quoting screen or hand an AI agent open access to the revenue system and hope it gets the pricing right, and neither one actually solves the problem. The real fix is to make quoting something a rep can do correctly from wherever they already are, with the same rules, the same discount limits, and the same approvals that apply inside Salesforce today. That's what Nue's MCP makes possible.

In this week's demo, a rep works from Claude Code, connected to Nue through its MCP servers. They make a single request via a voice prompt: create an opportunity and quote for the Acme Co. account, include the Vroom bundle and its required products, apply a 10% discount, and set a close date of tomorrow, leaving the name and stage to the assistant's judgment.

The rep doesn't click through a quote builder, hunt for the right screen, or pick which tools to use. The assistant decides that on its own: it looks up the bundle in the product catalog, creates the opportunity, and builds the quote. It then reports back in plain terms: the opportunity, the quote with its discount levels and bundle pricing, and the numbers that matter. The rep opens Salesforce afterward and finds the opportunity, the quote, and the exact quote lines requested, all sitting where they should be.

What used to take a dozen or more clicks through a quote editor now happens in one exchange. The speed is the obvious part. The part worth paying attention to is what didn't change: the pricing, the discount rules, and the approval logic behind that quote are identical to what a rep would get building it by hand.

Most tools enforce their pricing rules and discount limits through the UI itself. Take that screen away, and the rules go with it, which is exactly why letting an AI agent loose on a revenue system without that structure is risky: it can move fast and still get the deal wrong.

Nue doesn't work that way. Every request into Nue, whether it comes from a rep in the UI or an AI client, runs through the same pricing and approval logic and is tied to that person's own permissions, not a shared login everyone uses. A discount that would need approval in Salesforce needs approval here too. And the quote itself doesn't land in some separate holding area waiting to be reconciled later. Quotes, orders, subscriptions, usage, and billing all live on one system, so an agent-built quote shows up exactly the way a rep-built one would.

There are two common ways vendors are answering "AI plus quoting" right now, and both fall short. Some add an AI layer on top of the same old quoting screen, so the moment a request comes from anywhere else, none of the guardrails apply. Others go the opposite direction and expose their entire system to AI agents with no revenue-specific logic to keep them from mispricing a bundle or approving a discount they shouldn't.

Nue sits in between those two extremes on purpose. The rep, or the agent working on their behalf, isn't given raw access to poke around a database. It's given the same governed actions, with the same pricing and approval logic built in, that already run every quote inside Nue today.

This demo isn't about the AI client. It's about a rep getting real work done from wherever they already are, while the quote and the underlying data land in one auditable source of truth, so the financial data flowing downstream stays clean.

The same MCP connection powering this demo could just as well power a streamlined internal tool, like a simple app where a CSM types in an account name and gets a renewal quote back, or a Slack workflow where reps and CSMs build quotes and renewals from a message. And it isn't limited to quoting, either. The same governed engine extends to pricing, renewals, billing, and collections: a rep could ask the same assistant to update a price on an existing order, or a CSM could ask it to check a customer's invoice status, with the same guardrails applying either way. This isn't a feature added onto a quoting tool. It's the architecture the entire revenue process runs on.

The AI client is the surface. Nue is what's actually doing the work underneath it.