Skip to main content
This section covers the essential building blocks for creating MCP servers with FrontMCP. Whether you’re building a simple tool wrapper or a complex multi-app system, understanding these core concepts is key.

Core Components

The building blocks of every FrontMCP server. These components define what your server exposes to AI clients.

Server

Entry point, configuration, and composition modes for your MCP server

Apps

Container for grouping related tools, resources, and prompts with shared auth

Tools

AI-callable functions with typed input/output schemas and validation

Resources

Data endpoints for static content and dynamic templates

Prompts

Reusable prompt templates with argument interpolation

Agents

LLM-powered autonomous agents that can use tools and make decisions

Elicitation

Request user input during tool or agent execution

Skills

HTTP-accessible instructions for Claude Code and other AI coding tools

Discovery

Server, app, and tool discovery for multi-server environments

Context Classes

Every tool, resource, prompt, agent, and skill runs within a context class. These provide access to dependency injection, logging, authentication, and more.

Context Overview

Introduction to context classes and the execution model

ExecutionContextBase

Base class with DI, logging, auth, and shared functionality

ToolContext

Context for tool execution with typed input/output

ResourceContext

Context for resource reads with URI parameters

PromptContext

Context for prompt generation with arguments

AgentContext

Context for agent execution with tool access

SkillContext

Context for skill execution via HTTP

Authentication

FrontMCP implements a flexible three-tier authentication system supporting public access, token pass-through, and full OAuth 2.1 orchestration.
ModeDescriptionBest For
PublicNo auth required, anonymous sessionsDevelopment, testing, public APIs
TransparentPass-through tokens from external IdPsExisting Auth0, Okta, Azure AD setups
OrchestratedFull OAuth 2.1 server (local or remote)Multi-app, federated auth, progressive auth

Auth Overview

Understanding the three-tier authentication system

Authorization Modes

Deep dive into public, transparent, and orchestrated modes

Local OAuth

Self-contained OAuth server with built-in user management

Remote OAuth

Proxy to external identity providers (Auth0, Okta, etc.)

Progressive Auth

Request additional scopes during tool execution

CIMD

Client-Initiated Mutual Disambiguation for multi-account scenarios

Deployment

Options for running your FrontMCP server in development and production.
TargetUse CaseSession Storage
Local DevDevelopment and testingIn-memory
Production BuildLong-running serverRedis or Vercel KV
ServerlessEdge/Lambda deploymentsExternal storage required

Local Development

Start the development server with hot reload

Production Build

Build and run optimized production bundles

Redis Setup

Configure Redis for session and token storage

Vercel KV

Use Vercel KV as a Redis alternative

Serverless

Deploy to serverless platforms (Vercel, AWS Lambda)

Where to Start

New to FrontMCP? Start with Server to understand the entry point, then explore Tools to see how AI-callable functions work. Once you’re comfortable with the basics, dive into Authentication for securing your server.

Quickstart

Build your first FrontMCP server in minutes

SDK Reference

Full API documentation for decorators, classes, and registries