# AgentFront: FrontMCP v1.4 > AI Agent Infrastructure - MCP servers, secure sandboxes, vector databases, and more... ## v1.4 - [FrontMCP / v1.4 / Documentation (139 pages)](https://docs.agentfront.dev/_llms/front-mcp/v1-4/documentation.md): Documentation for FrontMCP / v1.4 / Documentation. ### SDK Reference #### Decorators - [Decorators Overview](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/decorators/overview.md): FrontMCP uses TypeScript decorators to define MCP server components declaratively. All decorators follow a metadata-first design with Zod schema validation. - [@FrontMcp](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/decorators/frontmcp.md): The @FrontMcp decorator is the entry point for creating an MCP server. It configures the server, registers apps, and manages the server lifecycle. - [@App](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/decorators/app.md): The @App decorator defines an application module that groups related tools, resources, prompts, agents, and other components. - [@Tool](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/decorators/tool.md): The @Tool decorator defines an MCP tool with validated input/output schemas and full TypeScript type inference. - [@Resource](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/decorators/resource.md): The @Resource decorator defines a static MCP resource with a fixed URI that can be read by AI clients. - [@ResourceTemplate](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/decorators/resource-template.md): The @ResourceTemplate decorator defines a dynamic MCP resource with a URI template (RFC 6570) that extracts parameters from the requested URI. - [@Prompt](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/decorators/prompt.md): The @Prompt decorator defines an MCP prompt template that can be invoked by AI clients to generate structured prompts. - [@Agent](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/decorators/agent.md): The @Agent decorator defines an autonomous AI agent that uses an LLM to execute tasks with access to tools. - [@Skill](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/decorators/skill.md): The @Skill decorator defines a knowledge package that bundles instructions and tool references for specific workflows. - [@Provider](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/decorators/provider.md): The @Provider decorator registers a class as a dependency injection provider that can be injected into tools, resources, agents, and other components. - [@Plugin](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/decorators/plugin.md): The @Plugin decorator creates reusable modules that bundle providers, tools, resources, and context extensions. - [@Job](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/decorators/job.md): The @Job decorator defines an executable job with validated input/output schemas, retry configuration, and permission checks. - [@Workflow](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/decorators/workflow.md): The @Workflow decorator defines a DAG-based multi-step pipeline that composes jobs with dependencies, conditions, and parallel execution. - [@Channel](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/decorators/channel.md): The @Channel decorator defines a push-based notification channel with source configuration and optional two-way communication. - [Flow Hooks](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/decorators/hooks.md): FrontMCP provides hook decorators for intercepting and extending flow execution. Hooks allow you to add cross-cutting concerns like logging, validation, and caching. #### API ##### Core Classes - [FrontMcpInstance](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/core/frontmcp-instance.md): FrontMcpInstance is the main entry point for creating and managing FrontMCP servers. It provides multiple factory methods for different deployment scenarios. - [Scope](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/core/scope.md): Scope provides access to all registries and services within a FrontMCP server. It's the central point for accessing tools, resources, prompts, agents, skills, and other registered components. - [FrontMcpServer](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/core/server.md): FrontMcpServer manages the HTTP server interface for FrontMCP. It supports multiple host adapters (Express, Fastify, etc.) and both server and serverless deployments. - [DirectClient](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/core/direct-client.md): DirectClient provides programmatic access to FrontMCP servers with automatic tool formatting for different LLM platforms (OpenAI, Claude, Vercel AI, etc.). ##### Context Classes - [Context Classes Overview](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/contexts/overview.md): FrontMCP provides context classes that serve as the base for all entry types (tools, resources, prompts, agents, skills). These classes provide access to dependency injection, logging, authentication, and other services. - [ExecutionContextBase](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/contexts/execution-context-base.md): ExecutionContextBase is the abstract base class that provides common functionality for all entry context classes (ToolContext, ResourceContext, AgentContext, SkillContext). - [ToolContext](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/contexts/tool-context.md): ToolContext is the base class for all tool implementations. It extends ExecutionContextBase and provides tool-specific features like notifications, progress reporting, and elicitation. - [ResourceContext](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/contexts/resource-context.md): ResourceContext is the base class for all resource implementations. It extends ExecutionContextBase and provides resource-specific features like URI handling and parameter extraction. - [PromptContext](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/contexts/prompt-context.md): PromptContext is the base class for all prompt implementations. It provides access to prompt arguments and methods for building prompt responses. - [AgentContext](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/contexts/agent-context.md): AgentContext is the base class for autonomous AI agents. It extends ExecutionContextBase and provides LLM integration, tool execution, and the default agent loop. - [SkillContext](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/contexts/skill-context.md): SkillContext is the base class for skill implementations. It extends ExecutionContextBase and provides methods for loading instructions and building skill content. - [JobContext](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/contexts/job-context.md): JobContext is the base class for all job implementations. It extends ExecutionContextBase and provides job-specific features like logging, progress reporting, and retry tracking. - [ChannelContext](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/contexts/channel-context.md): Abstract base class for channel execution contexts. Handles incoming events and transforms them into Claude Code notifications. ##### Registries - [Registries Overview](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/registries/overview.md): FrontMCP uses registries to manage collections of tools, resources, prompts, agents, skills, and other components. All registries follow common patterns for registration, lookup, and change events. - [ToolRegistry](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/registries/tool-registry.md): The ToolRegistry manages all MCP tools within a scope. It provides methods for listing, finding, and subscribing to tool changes. - [ResourceRegistry](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/registries/resource-registry.md): The ResourceRegistry manages all MCP resources within a scope, including both static resources and resource templates. - [PromptRegistry](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/registries/prompt-registry.md): The PromptRegistry manages all MCP prompts within a scope. Prompts are reusable message templates that can be parameterized. - [AgentRegistry](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/registries/agent-registry.md): The AgentRegistry manages all MCP agents within a scope. Agents are autonomous components that can be called as tools and have their own set of visible tools. - [SkillRegistry](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/registries/skill-registry.md): The SkillRegistry manages all MCP skills within a scope. Skills are reusable prompts/instructions that can be searched, loaded, and optionally synced to external storage. - [PluginRegistry](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/registries/plugin-registry.md): The PluginRegistry manages all plugins within a scope. Plugins extend FrontMCP functionality by providing tools, resources, prompts, skills, providers, and hooks. - [HookRegistry](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/registries/hook-registry.md): The HookRegistry manages all flow hooks within a scope. Hooks intercept and extend the execution of tools, resources, and prompts at various stages. - [ProviderRegistry](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/registries/provider-registry.md): The ProviderRegistry manages dependency injection providers within a scope. It handles provider instantiation, scoping, and lifecycle management. - [AuthRegistry](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/registries/auth-registry.md): The AuthRegistry manages authentication providers within a scope. It handles auth configuration, provider detection, and orchestration. - [JobRegistry](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/registries/job-registry.md): The JobRegistry manages all jobs within a scope. It provides methods for listing, finding, searching, and subscribing to job changes. - [WorkflowRegistry](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/registries/workflow-registry.md): The WorkflowRegistry manages all workflows within a scope. It provides methods for listing, finding, searching, and subscribing to workflow changes. - [ChannelRegistry](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/registries/channel-registry.md): Registry that manages channel instances, advertises capabilities, and provides lookup methods. ##### APIs - [Guard API Reference](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/guard.md): Complete API reference for @frontmcp/guard — rate limiting, concurrency, timeout, and IP filtering. - [Telemetry API](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/telemetry.md): API reference for this.telemetry — custom spans, events, and attributes in execution contexts #### Errors - [Errors Overview](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/errors/overview.md): FrontMCP provides a comprehensive error hierarchy with MCP-compliant error codes for proper JSON-RPC error responses. - [Tool & Prompt Errors](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/errors/tool-errors.md): Errors thrown during tool and prompt lookup and execution. - [Resource Errors](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/errors/resource-errors.md): Errors thrown during resource lookup, reading, and URI validation. - [Validation Errors](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/errors/validation-errors.md): Errors for input/output validation failures and method mismatches. - [Auth Errors](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/errors/auth-errors.md): Errors for authentication failures, session issues, and authorization requirements. - [Agent Errors](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/errors/agent-errors.md): Errors for agent lookup, execution, configuration, and LLM adapter failures. - [Rate Limit Errors](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/errors/rate-limit-errors.md): Errors for rate limiting and quota enforcement. - [Transport Errors](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/errors/transport-errors.md): Errors for transport layer failures including connection, session, and content type issues. - [Remote Errors](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/errors/remote-errors.md): Errors for remote MCP server connections, operations, and configuration. - [Elicitation Errors](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/errors/elicitation-errors.md): Errors for MCP elicitation requests, timeouts, and store issues. - [Provider Errors](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/errors/provider-errors.md): Errors for dependency injection, provider resolution, and scope management. - [Registry Errors](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/errors/registry-errors.md): Errors for registry lookups, dependency resolution, and entry validation. - [Decorator Errors](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/errors/decorator-errors.md): Errors for invalid decorator metadata and hook target definitions. - [Normalization Errors](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/errors/normalization-errors.md): Errors for invalid provider and entity definitions during normalization. - [SDK Errors](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/errors/sdk-errors.md): General SDK errors for flows, configuration, skills, serverless, and runtime issues. - [Auth Internal Errors](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/errors/auth-internal-errors.md): Internal errors for encryption, vaults, token management, and session infrastructure. - [ESM Errors](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/errors/esm-errors.md): Error classes for ESM package loading, version resolution, manifest validation, caching, and registry authentication - [Task Errors](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/errors/task-errors.md): Error classes for the task-augmented tool invocation lifecycle - [Workflow Errors](https://docs.agentfront.dev/frontmcp/v/1.4/sdk-reference/errors/workflow-errors.md): Error classes for multi-step workflow execution ### Nx Plugin #### Get Started - [Nx Plugin Overview](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/overview.md): @frontmcp/nx — Nx generators and executors for FrontMCP monorepos - [Monorepo Patterns](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/guides/monorepo-patterns.md): Architecture patterns for large FrontMCP monorepos - [Migration from Standalone](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/guides/migration-from-standalone.md): Migrate a standalone FrontMCP project to an Nx monorepo #### Generators - [Generators Overview](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/generators/overview.md): All 20 FrontMCP Nx generators for scaffolding projects and components - [Workspace Generator](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/generators/workspace.md): Scaffold a full FrontMCP Nx monorepo with apps/, libs/, and servers/ directories - [App Generator](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/generators/app.md): Generate a FrontMCP application in apps/ - [Library Generator](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/generators/lib.md): Generate a shared library in libs/ - [Server Generator](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/generators/server.md): Generate a deployment shell in servers/ - [Tool Generator](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/generators/tool.md): Generate a @Tool class - [Resource Generator](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/generators/resource.md): Generate a @Resource or @ResourceTemplate class - [Prompt Generator](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/generators/prompt.md): Generate a @Prompt class - [Skill Generator](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/generators/skill.md): Generate a @Skill class - [Skill Directory Generator](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/generators/skill-dir.md): Generate a SKILL.md-based skill directory bundle - [Agent Generator](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/generators/agent.md): Generate an @Agent class - [Provider Generator](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/generators/provider.md): Generate a @Provider class for dependency injection - [Plugin Generator](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/generators/plugin.md): Generate a @Plugin class extending DynamicPlugin - [Adapter Generator](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/generators/adapter.md): Generate an @Adapter class extending DynamicAdapter - [Auth Provider Generator](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/generators/auth-provider.md): Generate an @AuthProvider class - [Flow Generator](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/generators/flow.md): Generate a @Flow class extending FlowBase - [Job Generator](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/generators/job.md): Generate a @Job class - [Workflow Generator](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/generators/workflow.md): Generate a @Workflow class - [UI Component Generator](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/generators/ui-component.md): Add a reusable MUI-based React component to ui/components/ - [UI Page Generator](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/generators/ui-page.md): Add a full-page React component to ui/pages/ - [UI Shell Generator](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/generators/ui-shell.md): Add a server-side HTML shell template to ui/shells/ #### Executors - [Executors Overview](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/executors/overview.md): All 7 FrontMCP Nx executors for building, testing, and deploying - [Build Executor](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/executors/build.md): Compile a FrontMCP project using frontmcp build - [Build Exec Executor](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/executors/build-exec.md): Build a distributable bundle using frontmcp build --target node - [Dev Executor](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/executors/dev.md): Start a development server using frontmcp dev - [Serve Executor](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/executors/serve.md): Start a supervised production server using frontmcp start - [Test Executor](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/executors/test.md): Run E2E tests using frontmcp test - [Inspector Executor](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/executors/inspector.md): Launch MCP Inspector using frontmcp inspector - [Deploy Executor](https://docs.agentfront.dev/frontmcp/v/1.4/nx-plugin/executors/deploy.md): Deploy to a target platform (node/vercel/lambda/cloudflare)