Skip to content

MCP Gateway — Overview

Developer

The MCP Gateway is a Prodgy component that exposes your organization’s integrations, knowledge base, and AI agents as MCP tools — the standard protocol used by modern AI coding assistants.

This means tools like Claude Code, Cursor, Windsurf, and Claude Desktop can directly interact with your Prodgy workspace: creating work items, querying your knowledge base, executing agents, and managing integrations — all from within the developer’s IDE.


These tools are always available:

ToolDescription
get_my_profileReturns the authenticated user identity, organization metadata and personal AI preferences. Auto-injected on session start.
get_workspace_infoReturns workspace context — vision, target audience, OKRs, active integrations. Auto-injected on session start.
update_my_ai_preferencesUpdates the user’s personal AI directive (markdown) and the destructive-action confirmation flag
search_knowledgeSemantic search (RAG) on the workspace’s knowledge base
list_agentsList AI agents configured for the workspace
invoke_agentExecute a Prodgy agent (with its own LLM, tools, and system prompt)
list_integration_catalogList available integration templates
get_integrationGet details of an integration template
list_installed_integrationsList integrations currently installed in the organization
install_integrationInstall an integration into the organization with credentials
create_integrationCreate an integration template (SUPERADMIN/ADMIN role)
update_integrationUpdate an integration template (SUPERADMIN/ADMIN role)
delete_integrationDelete an integration template (SUPERADMIN/ADMIN role)
get_active_integrationsList active integrations for the organization
get_mcp_settingsGet MCP Server configuration (SUPERADMIN/ADMIN role)
update_mcp_settingsUpdate MCP Server instructions and rules (SUPERADMIN/ADMIN role)

Beyond the platform tools, the MCP Gateway dynamically exposes tools from your organization’s active integrations. Each integration you configure in Prodgy (Azure DevOps, Jira, Slack, GitHub, etc.) automatically adds its own tools to the Gateway — no code changes needed.

This means the set of available tools grows as you add integrations, and your AI assistant can interact with all of them through a single MCP connection.


The MCP Gateway supports customizable instructions that define how AI clients should behave when interacting with your workspace. You can manage these instructions through:

  1. Prodgy UI: Settings > MCP Server
  2. MCP tool: call update_mcp_settings (requires SUPERADMIN role)

The instructions allow you to define behaviors such as always loading the project context first, how to handle integration errors, and rules for creating new integration templates.


The MCP Gateway supports multiple credential patterns for integrations, all resolved server-side:

  • Header-based — e.g., Bearer {token}
  • Query parameter — credentials injected into endpoint URLs
  • Request body — credentials injected into request bodies
  • OAuth2 client_credentials — automatic token exchange before each request
  • Multi-step session — login + session token acquisition before each request

  • How It Works — tool discovery, execution flow, and access control
  • How to Use — step-by-step setup for Claude Code, Cursor, Windsurf, and Claude Desktop