MCP server · Campus Dyno Docs

Model Context Protocol surface so MCP-native agents plug in without writing a vendor-specific connector.

What MCP is

Anthropic’s Model Context Protocol (MCP) is an open standard for connecting AI agents to tools and data sources. An MCP server exposes a set of tools, resources, and prompts. An MCP-aware agent (Claude desktop, custom LangChain/CrewAI agents, future Claude apps) discovers and uses them without per-vendor integration code.

For Campus Dyno, MCP is the answer to the question “how does my custom Slack agent talk to your CPQ?” when the answer “write a REST client against our OpenAPI spec” is too heavy.

Planned tool surface

The MCP server will mirror the public REST API as a set of named tools:

quote.create: create a new quote from a structured intent object
quote.get: fetch a quote by ID
quote.update_line_items: add/remove/edit line items
quote.request_approval: submit for approval (routing follows the on-behalf-of-user’s scope)
quote.send: generate document, issue signature link
catalog.list: list products
catalog.get: fetch product detail
audit.query: query the audit log
audit.replay: replay a past action for reasoning about historical state

Auth

The MCP server uses the same agent-on-behalf-of-user token model as the REST API. An agent connecting via MCP authenticates as a specific deal-desk user, and every tool call is recorded in the audit log with the (agent, on-behalf-of-user) pair.

Why this matters

The agent ecosystem is splitting along an MCP / not-MCP line. Claude desktop, the Anthropic SDK, and a growing set of custom agent stacks speak MCP natively. The vendors that don’t ship MCP servers (Salesforce CPQ’s closed data model, PandaDoc’s thin API, HubSpot Quotes’ feature-not-product surface) will be paired with custom middleware. Campus Dyno’s MCP surface means a customer running Claude as their agent layer doesn’t write any glue code.

What you can do today

The REST API (/docs/api) is the path until the MCP server reaches beta. If you’re building an MCP-native agent and Campus Dyno would be a natural tool surface for you, start a conversation. Design-partner customers pull the MCP work forward.