Topics Covered:

Skills Overview: How Workers acquire and use capabilities
API Integrations: Using Providers and Connectors as skills
AI Workflows as Skills: Embedding workflows within AI Workers
MCP (Model Context Protocol): Connecting to external tool servers

Skills Overview

Skills are the capabilities that enable Workers to interact with external systems, execute specialized logic, and perform actions beyond basic conversation. Skills transform AI Workers from simple chat interfaces into powerful automation agents.

Types of Skills:

API Integrations: External services accessed via Providers and Connectors (e.g., Microsoft 365, Salesforce, GitHub)
AI Workflows: Canvas-built workflows embedded as callable tools within AI Workers
MCP Tools: Tools discovered from external MCP (Model Context Protocol) servers

How Skills Work:

Configuration: Builders add skills to Workers in the AI Worker Builder's Skills tab
Discovery: The Worker learns what actions are available through skill definitions
Invocation: During conversations, the LLM decides when to use a skill based on user intent
Execution: The platform executes the skill and returns results to the conversation
Session Control: Users can enable/disable specific skills per session

API Integrations

API integrations allow Workers to interact with external services through Providers and Connectors.

Adding API Skills:

Navigate to the Skills tab in the AI Worker Builder
Select from available Connectors (configured in the Provider system)
Choose specific operations (endpoints) to expose as skills
Configure any default parameters or constraints

Skill Behavior:

Workers automatically understand when to call API skills based on conversation context
Authentication is handled transparently via Connectors
Results are formatted and presented naturally in the conversation
Failed calls are handled gracefully with appropriate error messages

Best Practices:

Only add skills that are relevant to the Worker's purpose
Use descriptive operation names to help the LLM understand when to use them
Configure appropriate rate limits and timeouts
Test skills thoroughly before deploying to users

AI Workflows as Skills

AI Workflows built in Canvas can be embedded as skills within AI Workers, enabling complex multi-step operations to be triggered from natural conversation.

Benefits:

Encapsulation: Complex logic is packaged into a single callable skill
Reusability: One workflow can be used across multiple Workers
Maintainability: Update the workflow once, all Workers using it get the update
Separation of Concerns: Business logic lives in workflows, conversation logic in Workers

Configuring Workflow Skills:

Build and test your AI Workflow in Canvas
In the AI Worker Builder, go to the Skills tab
Select "Add AI Workflow" and choose from available workflows
Map input/output parameters as needed
Provide a clear description to help the LLM understand when to invoke it

Execution Flow:

User makes a request in chat
LLM determines the workflow skill is needed
Worker extracts parameters from the conversation
Workflow executes with those parameters
Results return to the conversation

MCP (Model Context Protocol)

Model Context Protocol (MCP) is an open standard that enables Workers to discover and use tools from external servers. MCP provides a standardized way to extend Worker capabilities without building custom integrations.

What is MCP?

MCP defines how AI systems can:

Discover available tools, prompts, and resources from external servers
Invoke tools with structured parameters
Receive results in a consistent format

This allows EverWorker to connect to any MCP-compatible server and immediately gain access to its capabilities.

Key Concepts:

Concept	Description
MCP Server	An external service that exposes tools via the MCP protocol
MCP Tools	Individual capabilities exposed by a server (e.g., "search_database", "send_email")
MCP Prompts	Pre-defined prompt templates provided by the server
MCP Resources	Data or content URIs that can be accessed via the server

Connecting to MCP Servers

Builders can connect to external MCP servers to expand Worker capabilities:

Navigate to MCP Settings: Access MCP server configuration in the platform
Add Server: Provide the server URL and authentication details
Sync Tools: Discover available tools, prompts, and resources
Assign to Workers: Add MCP tools as skills in AI Workers

Authentication Options:

MCP servers can be configured with different authentication methods:

None: For public or localhost servers
Bearer Token: Standard token-based authentication
API Key: Key-based authentication via headers

Using MCP Tools in Workers:

Once an MCP server is connected and synced:

Go to the Skills tab in AI Worker Builder
Select "Add MCP Tools"
Browse available tools from connected servers
Select tools to enable for this Worker
Tools appear as callable skills during conversations

Multi-Provider Compatibility:

EverWorker automatically converts MCP tool definitions to work with different LLM providers:

OpenAI: Converted to function calling format
Anthropic: Converted to Claude tool format
Google: Converted to Gemini function declarations

This means the same MCP tools work regardless of which LLM powers your Worker.

Internal MCP Server:

EverWorker also exposes its own capabilities via an internal MCP endpoint (/mcp), allowing external systems to interact with the platform using the MCP protocol.

Best Practices for MCP:

Verify Server Trust: Only connect to MCP servers you trust
Review Tool Permissions: Understand what each tool can do before enabling
Monitor Usage: Track MCP tool invocations for security and debugging
Keep Synced: Periodically re-sync to discover new or updated tools

Skill Management

Session-Level Control:

Users can enable or disable skills per session, allowing them to:

Focus the Worker on specific tasks
Prevent certain actions when not needed
Troubleshoot by isolating skill behavior

Visibility:

When chatting with a Worker, users can view:

Which skills are currently enabled
What each skill does (descriptions)
When skills are being invoked (thinking trace)

Permissions:

Builders configure which skills are available to a Worker
Users can toggle enabled skills within their session
Admins can restrict certain skills organization-wide

Summary

Skills are what transform AI Workers from conversational assistants into powerful automation tools. Whether through API integrations, embedded AI Workflows, or MCP-connected external tools, skills enable Workers to take action in the real world. The modular skill system ensures that capabilities can be added, updated, and managed independently - keeping Workers flexible, maintainable, and secure.