Skip to Content

Key concepts of Jitterbit MCP

This page defines the components and concepts you'll encounter while working with Jitterbit MCP.

Model Context Protocol

The Model Context Protocol (MCP) is an open protocol that lets AI systems discover and use capabilities exposed by external systems in a structured, standardized way. An MCP server can expose three types of capabilities:

  • Tools: Actions that clients can invoke.

  • Resources: Read-only data and contextual information that clients can access.

  • Prompts: Reusable prompt templates and guidance that clients can use.

Jitterbit MCP supports the 2026-07-28 and 2025-11-25 MCP protocol versions. The MCP gateway automatically detects which version a connecting client uses, so no client-side configuration is required.

MCP control plane

The MCP control plane provides centralized configuration, management, and governance for Jitterbit MCP. It manages:

  • MCP servers, tools, resources, and prompts
  • Environment and deployment configuration
  • Access policies
  • Authentication and service tokens
  • Publishing and lifecycle management
  • Analytics and monitoring
  • Auditing and compliance information

The control plane manages configuration and operational metadata. It does not directly process MCP client requests.

MCP runtime

The MCP runtime hosts MCP servers and processes requests for the capabilities they expose. The runtime is responsible for:

  • Advertising the capabilities an MCP server supports.
  • Processing tool, resource, and prompt requests.
  • Validating requests against published definitions.
  • Invoking the underlying integrations and APIs used by tools.
  • Returning results through the MCP gateway.

The MCP runtime is available in two deployment models:

  • Cloud MCP runtime: Hosted and managed by Jitterbit.

  • Private MCP runtime: Deployed and managed by you in a private or local environment. A private MCP runtime uses a runtime service token to authenticate with the MCP control plane.

MCP gateway

The MCP gateway provides the secure connection between MCP clients and MCP runtimes. It receives MCP traffic and can perform functions such as:

  • MCP protocol handling
  • Client authentication and authorization
  • Request routing
  • Policy enforcement
  • Rate limiting
  • Request and response logging
  • Analytics and observability

The MCP gateway is available in two deployment models:

  • Cloud MCP gateway: Hosted and managed by Jitterbit.

  • Private MCP gateway: Deployed and managed by you in a private or local environment. A private MCP gateway uses a gateway service token to authenticate with the MCP control plane.

How MCP traffic flows

When an MCP client invokes a tool, the request passes from the client through the gateway to the runtime, which invokes the underlying integration or API. The response returns along the reverse path, and usage information is recorded before the result reaches the client.

--- config: themeVariables: lineColor: '#6b7280' edgeLabelBackground: '#f5f5f5' flowchart: padding: 20 nodeSpacing: 80 --- flowchart TB classDef default fill:white, stroke:black, stroke-width:3px, rx:15px, ry:15px Client["
MCP clients
(AI assistants and agents)"] Gateway["
MCP gateway
(authenticates, applies policies, routes)"] Runtime["MCP runtime
(hosts MCP servers, invokes integrations)"] Assets["
Your integrations and APIs
(built in Studio and API Manager)"] Client -->|Invoke a tool| Gateway Gateway -->|Routes request| Runtime Runtime -->|Invokes| Assets Assets -->|Response| Runtime Runtime -->|Result| Gateway Gateway -->|Result| Client style Gateway fill:#eef2f7, stroke:#9fb3c8, stroke-width:2px

MCP server

An MCP server is a logical collection of capabilities exposed to MCP clients through a governed endpoint. An MCP server can contain:

  • Tools
  • Resources
  • Prompts
  • Authentication configuration
  • Environment and deployment settings

Publishing an MCP server makes its current published configuration available to authorized MCP clients. You create and manage MCP servers on the MCP Servers page.

MCP tool

An MCP tool is a discrete capability that an MCP client can invoke to perform an action. A tool can map to:

  • An external HTTP or REST API
  • A Jitterbit API created in API Manager
  • A Studio operation, triggered directly without a published Jitterbit API

Example tools include retrieving customer details, creating an order, submitting an invoice, sending a notification, and starting an integration operation.

Each tool is defined by structured metadata that can include its name and description, input schema, output schema, authentication configuration, and API or integration mapping. Tool definitions and schemas are validated when the MCP server is published.

MCP resource

An MCP resource provides read-only data or contextual information that an MCP client can access through an MCP server. Examples include customer profiles, contract information, product documentation, usage metrics, and reference data.

A resource's content can come directly from its own configuration, or be retrieved from an external HTTP or REST API, a Jitterbit API, or a file. For the configuration fields available when you create a resource, see Resource configuration.

MCP prompt

An MCP prompt is a reusable prompt template that helps MCP clients perform a defined task consistently. A prompt can include instructions, input arguments, context templates, and recommended workflows. Examples include a customer renewal summary, a support case triage, and a marketing campaign brief. For the configuration fields available when you create a prompt, see Prompt configuration.

MCP client

An MCP client is an application or system that uses the MCP protocol to discover and access the capabilities exposed by an MCP server. Examples include AI assistants, AI agents, agent orchestration platforms, development tools, and MCP testing clients.

An MCP client connects through the MCP gateway and can access only the MCP servers and capabilities for which it is authorized.

Governance and policy enforcement

Jitterbit MCP applies governance controls to MCP traffic and to the capabilities exposed by MCP servers. Governance capabilities can include:

  • Server-level and tool-level access control
  • Authentication and authorization
  • Rate limiting
  • Request and payload size limits
  • Tool enable and disable controls
  • Request routing policies
  • Logging and auditing
  • Monitoring and analytics
  • Protection against unsafe or malicious requests

Policies are managed through the MCP control plane and enforced by the MCP gateway and MCP runtime, as applicable.

Credentials and secret management

Jitterbit MCP securely manages the credentials required to invoke the underlying integrations and APIs. When a tool is invoked, the configured credentials are retrieved and applied during request processing. The credentials are not returned to or exposed to the MCP client.

Credential management can include secure credential storage, authentication configuration, credential injection, token management, and credential rotation.

Authentication tokens and service tokens

Jitterbit MCP uses different token types for different authentication purposes:

  • MCP authentication token: Authenticates an MCP client with an individual MCP server. Manage these on the Auth Tokens tab of the MCP Servers page.

  • Runtime service token: Authenticates a private MCP runtime with the MCP control plane.

  • Gateway service token: Authenticates a private MCP gateway with the MCP control plane.

Service tokens are required only for the corresponding private deployment components. Manage them on the MCP Service Tokens page.

Analytics and observability

Jitterbit MCP captures usage and operational metrics for MCP requests, including invocation counts, success and failure rates, request latency, usage by MCP server, usage by tool, and request volume over time. You can view the supported metrics on the Analytics page. To trace an individual MCP request and its related API and operation executions, see the Logs page.