Integration Builder

Developer

Tip

Installing an existing integration is available to all users. Creating new templates requires the SUPERADMIN or ADMIN role.

Overview

The MCP Gateway allows you to install and create integrations directly from your AI coding assistant — no need to leave your IDE.

• Install an existing integration — browse the catalog, provide your credentials and settings, and activate it immediately using install_integration.
• Create a new integration template — describe the service you want to connect, and Prodgy builds the complete template (requires SUPERADMIN role).

Once installed, the integration’s tools automatically appear in every team member’s MCP connection.

---

Installing an Existing Integration

If the integration you need already exists in the Prodgy catalog, you can install it directly from your IDE:

1. Browse the catalog

"List available integrations in Prodgy"

The AI will call list_integration_catalog to show you all available templates (Azure DevOps, Jira, Slack, GitHub, etc.).

2. Check required credentials and settings

"Show me the details of the Azure DevOps integration"

The AI will call get_integration to show you the required credential fields (e.g., personalAccessToken) and settings (e.g., project_url, project_name).

3. Install with your credentials

"Install the Azure DevOps integration with my PAT token xyz123 for https://dev.azure.com/myorg, project MyProject"

You can also specify a custom name and scope:

"Install the Azure DevOps integration as 'Azure DevOps — Backend' for this workspace only"

The AI will call install_integration which:

1. Validates your credentials against the external service (if the template has validation configured)
2. Creates the integration instance with the given name and scope (Local or Global)
3. Stores credentials securely (encrypted, server-side)
4. Activates the integration immediately — tools are available right away

You can install multiple instances of the same integration type, each with different credentials and scope — useful when different workspaces connect to different projects.

Tip

No need to go to the Prodgy UI! The entire flow — from browsing the catalog to having working tools — happens within your IDE.

4. Use the new tools

Once installed, the integration’s tools are immediately available:

"List open work items in Azure DevOps assigned to me"

"Create a GitHub issue for the authentication bug we just discussed"

---

Creating a New Integration Template

If the integration you need doesn’t exist in the catalog, you can create it (requires SUPERADMIN role).

1. Create the template

Ask your AI coding assistant to create an integration. For well-known services, it already knows the API structure:

"Create a Zendesk integration with tools for listing tickets, creating tickets, and updating ticket status"

The AI will use create_integration to build a complete template with:

• Authentication schema — how credentials are structured (API key, Bearer token, OAuth2, etc.)
• Configuration fields — what the admin fills in later (tenant URL, project name, subdomain, etc.)
• API tools — each action available as an MCP tool, with endpoints, parameters, and response mapping

Note

The template uses placeholders like ${settings.tenant_url} and ${credentials.apiKey}. No real credentials are needed at creation time — those are configured when installing.

2. Inspect and refine

Use list_integration_catalog to see all available templates, and get_integration to inspect a specific one:

"Show me the details of the Zendesk integration template"

Need to add more tools or fix something? Use update_integration:

"Add a tool to the Zendesk integration for searching tickets by status"

3. Install the new integration

After creating the template, install it immediately using install_integration:

"Install the Zendesk integration with my API key and subdomain"

Alternatively, the organization administrator can activate it via the Prodgy UI: Settings > Integrations.

---

Available Tools

Tool	Role	Description
list_integration_catalog	Any	List all available integration templates
get_integration	Any	Get full details of a specific template (credentials, settings, tools)
list_installed_integrations	Any	List integrations currently installed in the organization
install_integration	Any	Install a template with credentials, settings, optional title and product_id (for Local scope)
create_integration	SUPERADMIN/ADMIN	Create a new integration template
update_integration	SUPERADMIN/ADMIN	Update an existing template (add tools, fix configuration)
delete_integration	SUPERADMIN/ADMIN	Remove a template from the catalog

---

Best Practices

Be specific about the tools you need

Instead of “Create a Jira integration”, try:

"Create a Jira Cloud integration with tools for:
- Listing issues by project and status
- Creating new issues with summary, description, and priority
- Updating issue status
- Adding comments to issues"

Let the AI handle the API details

For well-known services (Jira, GitHub, Slack, Zendesk, Workday, etc.), the AI already knows the API structure. You only need to specify:

1. Which service to integrate
2. Which actions/tools you need
3. Which authentication type to use (if unsure, the AI will suggest)

Use generic tool names

Tool names should be generic, not prefixed with the integration name:

Correct	Incorrect
list_tickets	zendesk_list_tickets
create_issue	jira_create_issue
send_message	slack_send_message

This allows the same tool name to work across multiple integrations with automatic disambiguation.

Iterate incrementally

Start with the most important tools, activate the integration, test it, then add more tools as needed. It’s easier to build incrementally than to get everything right in one shot.