Smithery

What is Smithery?

Smithery is a marketplace and management platform for Model Context Protocol (MCP) servers, giving AI agents and large language models a central place to discover, connect to, and use external tools, APIs, and services. It functions as a registry where developers can browse thousands of community-built MCP servers, from web search and GitHub integrations to databases and cloud APIs, without building custom connectors from scratch. The platform targets AI agent developers, chatbot builders, and anyone extending tools like Claude Desktop or Cursor with external capabilities. Smithery describes its vision as building an "agent-first internet," where AI agents autonomously discover and call tools rather than relying on manual human-in-the-loop setups.

Key Features

• MCP Server Discovery and Management: Indexes over 7,600 MCP servers with one-click installation via CLI or web interface, covering services like GitHub, Google Sheets, Instagram, and Exa Search.
• Skills Marketplace: Provides over 131,000 community-built skills for tasks including PDF processing, financial modeling (DCF, Monte Carlo simulations), document generation, and web app testing via Playwright.
• Smithery Connect: A managed connection service offering zero OAuth setup, automatic token refresh, and secure credential storage for chatbots, agents, and copilots accessing MCP ecosystems.
• CLI Tool: The smithery-ai-cli package handles tool and skills discovery, schema inspection, and direct tool calls to services like Slack, Jira, Notion, and databases using commands like smithery tool list and smithery tool call.
• Developer Observability: Usage logs, analytics, benchmarks across LLMs, and testing frameworks that simulate LLM interactions, helping developers understand how their MCPs perform before shipping.
• Local and Hosted Server Options: Supports both local MCP servers where tokens stay on-device and hosted servers managed by Smithery with ephemeral data handling, giving developers control over their security posture.
• AI-Native UI Components: A set of components for chat interfaces, AI reasoning displays, workflow visualization, and code presentation that auto-activate on relevant queries.

Use Cases

• AI agent developers building chatbots or copilots: Developers connecting agents to external services like GitHub, Slack, or Google Sheets can install MCP servers in minutes using CLI commands, getting rapid integrations without writing custom API wrappers.
• MCP server vendors and distributors: Teams shipping their own MCP servers can use Smithery's CLI scaffolding (npm create smithery), hot-reloading dev environment, automatic GitHub deployment, and observability tools to test and distribute their servers to the broader agent ecosystem.
• Automation tool creators: Builders integrating agents with databases, project management tools, or blockchain data can use the registry to find purpose-built MCP servers, such as the Blockscout MCP, and connect them via simple CLI commands.
• Developers extending AI coding tools: Users of tools like Cursor or Claude Desktop can browse the registry to find MCP servers that add capabilities like web search, document processing, or API access directly to their existing AI assistant setup.

Strengths and Weaknesses

Strengths:

• Easy discovery and installation of MCP tools, with a large and growing catalog of servers and skills.
• Local-first approach lets developers keep sensitive tokens on-device, reducing exposure.
• Addresses a real integration pain point by centralizing MCP server access rather than requiring custom setups per service.
• Active development with CLI tooling, benchmarking, and observability features built specifically for agent workflows.

Weaknesses:

• A path traversal vulnerability in the Docker build process previously exposed around 3,000 hosted MCP servers to potential compromise, including API keys and secrets. Smithery patched the issue within 48 hours, but the incident highlighted risks in the hosted model.
• Over 20 open GitHub issues in the CLI repository report deployment failures with specific MCP servers, Windows-specific errors, and authentication bugs.
• Compatibility problems have been noted with tools like Cursor, where MCP servers from the registry sometimes fail to work as expected.

Pricing

• Hobby: Free, includes 25,000 RPCs per month, 3 namespaces, managed OAuth, and persistent connections.
• Pay as you Go: $30/month, includes a $30 monthly credit, usage billed at $0.50 per 1,000 RPCs beyond the credit, 100 namespaces, managed OAuth, and persistent connections.
• Custom: Contact Smithery for pricing. Includes everything in Pay as you Go plus custom rate limits, an uptime SLA, and Slack support.

Publishing an MCP server to the registry is free. The Hobby tier covers basic hosting, observability, and an MCP landing page at no cost.

FAQ

What is Smithery?

Smithery is a marketplace and management platform for Model Context Protocol (MCP) servers, giving AI agents and large language models a central place to discover, connect to, and use external tools, APIs, and services. It indexes over 7,600 MCP servers and more than 131,000 community-built skills.

What does Smithery mean?

In this context, Smithery is the name of a platform for discovering and managing MCP servers. It is not used in its traditional sense referring to a metalworking trade or workshop.

What is an MCP server?

An MCP server is an external tool, API, or service that AI agents and large language models can connect to and call. Smithery's registry includes MCP servers for services like GitHub, Slack, Google Sheets, Notion, and databases.

How does Smithery work?

Smithery provides a registry where developers can browse MCP servers and install them via a CLI tool or web interface. Commands like smithery tool list and smithery tool call handle discovery, schema inspection, and direct tool calls.

What is Smithery Connect?

Smithery Connect is a managed connection service that offers zero OAuth setup, automatic token refresh, and secure credential storage. It is designed for chatbots, agents, and copilots accessing MCP ecosystems.

What is the Smithery CLI?

The smithery-ai-cli package is a command-line tool for discovering tools and skills, inspecting schemas, and calling services like Slack, Jira, Notion, and databases directly. It supports one-click installation of MCP servers.

What is the difference between Smithery and Composio?

Smithery is specifically a marketplace and management platform built around the Model Context Protocol, indexing MCP servers and community skills. The provided documentation does not include a direct comparison to Composio.

Who is Smithery built for?

Smithery targets AI agent developers, chatbot builders, and anyone extending tools like Claude Desktop or Cursor with external capabilities. It also serves MCP server vendors who want to distribute and test their own servers.

What skills does Smithery offer?

Smithery provides over 131,000 community-built skills covering tasks including PDF processing, financial modeling, document generation, and web app testing via Playwright.

Does Smithery store credentials or tokens?

Smithery offers both local MCP servers, where tokens stay on-device, and hosted servers managed by Smithery with ephemeral data handling. Developers can choose their preferred security posture based on these options.

Can developers publish their own MCP servers on Smithery?

Yes. Smithery provides CLI scaffolding via npm create smithery, a hot-reloading development environment, automatic GitHub deployment, and observability tools for teams shipping their own MCP servers.

What observability tools does Smithery provide?

Smithery includes usage logs, analytics, benchmarks across LLMs, and testing frameworks that simulate LLM interactions. These tools help developers understand how their MCP servers perform before shipping.

What is Smithery's stated vision?

Smithery describes its vision as building an "agent-first internet," where AI agents autonomously discover and call tools rather than relying on manual human-in-the-loop setups.