Value teams have spent years working the value case inside one vendor's login while their sellers drafted somewhere else. Cuvama MCP ends that. Sellers, CSMs and value leaders can now read and update a live value plan from Claude, ChatGPT, Copilot or Gemini, or from agents their own team builds, with drivers and numbers staying inside the company's value model guardrails.
LONDON, Sept. 8, 2026 /PRNewswire-PRWeb/ -- Cuvama today made Cuvama MCP generally available, opening its value platform so any AI assistant or agent can work directly on a customer's value plans. Value teams can now work in whichever assistant their company has standardised on, or in agents they built themselves, and still work on the same governed value plan.
Value platforms have mostly been closed by design. The value case sits behind one vendor's login while the seller drafts in an assistant, the CSM works somewhere else, and the champion works in email and a spreadsheet nobody on the vendor side ever sees. Anything written outside the platform carries none of the company's value model and leaves no record of who agreed to what.
"Every value team we talk to is already building agents, and they were going to do that whether we helped or not," said MG Gurbaxani, co-founder & CEO of Cuvama. "The question we had to answer is whether the value case those agents produce is something a buyer's decision maker will accept. That depends entirely on what sits underneath the agent. So we opened the platform instead of asking people to choose between our interface and the one they already work in."
Cuvama MCP is an open interface to the Cuvama application, built on the Model Context Protocol. It splits into two surfaces. The User MCP is for sellers, value engineers and CSMs working a live deal: reading the value plan and the discovery notes, transcripts and CRM records behind it, turning what a call surfaced into a quantified value driver, building drivers and ROI inside the company's own value model guardrails, sharing the plan with a champion, and carrying it past the close so actuals sit next to what was promised. The Admin MCP is for value leaders: configuring the value model every plan inherits, setting templates and org settings, managing access, and reporting on adoption and results.
The interface is self-describing, which means an agent connecting to it inherits guidance as well as access. The first version it drafts follows the customer's methodology rather than a generic ROI template.
"An AI agent can create a basic value case in ninety seconds. That's not the hard part," said Alex Smith, co-founder & CPO of Cuvama. "What's important is whether the drivers are aligned with your value model, using your proven numbers, whether your champion helped build and own the value plan, and whether you have a living record that can be transferred to Customer Success and help continually inform your value story. That is what you need the platform is for, and our MCP provides the access your agents need to turn them into super-agents."
The company frames the release as a change in how value teams buy. "The build-versus-buy argument is changing," Gurbaxani said. "It is build and buy. You buy the governed value plan platform so your team can build whatever they want on top of it."
Cuvama MCP is available now to Cuvama customers. Connector setup guides are published for Claude, ChatGPT, Microsoft Copilot, Gemini Enterprise and Runlayer.
About Cuvama
Cuvama is an AI-native platform that turns early-stage sales discovery into defensible, executive-ready value cases. It sits between sales, strategy and finance, giving revenue teams one living value plan that carries a deal from value discovery through the value case to value realisation. Cuvama serves B2B technology companies with a dedicated value function, and every implementation includes practitioners who have built value programmes and carried a number themselves. Learn more at cuvama.com.
Media Contact
Patrick Moorhead, Cuvama Ltd., 1 3125436903, [email protected], https://cuvama.com
SOURCE Cuvama Ltd.

Share this article