Skip to main content

Prismatic's MCP Flow Server

What is MCP?​

MCP is an open protocol, introduced by Anthropic, that standardizes how LLMs request context from applications. When an LLM connects to an MCP server, it requests a list of tools that the MCP server offers.

These tools can perform actions such as "Find people with a specified name in HubSpot" or "Add an event to my Google Calendar." An MCP server offers a standardized schema so LLMs know to send a first and last name to the "find people" tool or an event name and date to the "add event" tool.

Why use Prismatic MCP?​

Out-of-the-box AI agents have limited capabilities - they can only access data provided to them and lack the authentication credentials or API endpoints needed to interact with the other apps and services your customers use. Without structured tooling, agents can also behave non-deterministically, calling wrong endpoints or mishandling errors.

Prismatic MCP solves this by exposing your existing integration flows as MCP tools. Your AI agent can interact with your customers' third-party apps in a deterministic, authenticated way - without custom API endpoints or additional code.

Example in-app chat using Prismatic MCP

How does Prismatic's MCP Flow Server work?​

You can designate some of your integrations' flows to be agent flows (i.e. MCP tools) that your AI agent can invoke. Using either the low-code designer or code-native SDK, you can assign an invocation schema to these agent flows, which tells the AI agent how to call them and what parameters to send.

When your AI agent queries the Prismatic MCP Flow Server, it gets a list of available tools - agent flows you've built - along with their invocation schema. If your customer wants to know How many contacts do I have in HubSpot named 'Bill'?, the AI agent can determine that it needs to call the Get contact count flow that you've built, extract the necessary parameters from the customer's query, and invoke that flow via the Prismatic MCP Flow Server.

Organization team members have access to agent flows of test instances

When querying for agent flows as an organization team member, you will receive a list of agent flows within your test instances (the instances you interact with in the integration designer). You will not see your customers' instances' agent flows.

Customer users, on the other hand, will see agent flows of production instances deployed to their customer.

How do I get started?​

  1. Build an agent flow by following the Getting Started guide.
  2. Test your agent flow using an MCP client, like Postman, ChatGPT or Claude.
  3. Connect your app's AI agent(s) to Prismatic's MCP server to give your customers access to your agent flows.