MCP Tools

Guide

Tip

When you install an integration, an MCP Tool is automatically registered in Prodgy, allowing AI agents to interact with the connected tool autonomously.

What is an MCP Tool?

MCP (Model Context Protocol) is the protocol used by Prodgy to connect AI agents to external tools. Each installed integration automatically registers an MCP Tool — a structured configuration that describes:

• Available actions — what the tool can do (e.g., create issue, send message, fetch data)
• Accepted parameters — what information the tool needs to perform each action
• Response formats — how results are returned

---

How It Works in Practice

When you chat in the Playground or interact with an agent, the MCP orchestrator analyzes your message and:

1. Identifies the intent — understands what you want to accomplish (e.g., “create a GitHub issue to fix bug X”)
2. Selects the appropriate MCP Tool — locates the corresponding integration among available tools
3. Executes the action — invokes the tool with the correct parameters automatically
4. Returns the result — presents the tool’s response directly in the conversation

Note

The entire process is transparent. You don’t need to configure anything beyond installing the integration — the MCP Tool is automatically registered and available for use by agents.

---

Usage Examples

Chat Intent	MCP Tool Triggered	Action Performed
”Create a GitHub issue”	GitHub	Creates issue in the configured repository
”Send a message on Slack”	Slack	Sends message to the specified channel
”List open PRs”	GitHub / GitLab	Queries open pull requests
”What’s the deploy status?”	Azure DevOps	Queries CI/CD pipeline
”Create a Jira task”	Jira	Creates task in the configured project

---

MCP Orchestrator

The MCP orchestrator is the component responsible for coordinating tool calls. It acts as an intermediary between the AI agent and installed integrations:

• Intelligent routing — directs each request to the correct MCP Tool based on conversation context
• Multiple tools — can trigger more than one tool in a single interaction if needed
• Preserved context — maintains conversation history for more precise calls
• Error handling — notifies the user if a tool is unavailable or a failure occurs

Caution

For an MCP Tool to work in chat, the corresponding integration must have an Active status with valid credentials. Integrations with Failed or Disconnected status will not be triggered by the orchestrator.

---

Relationship with Integrations

Aspect	Integration	MCP Tool
What it is	Configured connection to an external tool	AI invocation configuration
When created	When installing the integration	Automatically, along with the integration
Who uses it	Workflows, dashboards, agents	AI agents and MCP orchestrator
Configuration	Credentials and connection parameters	Automatic — no additional configuration needed