Smithery
Smithery is an MCP server registry and hosting platform that lets developers discover, publish, and connect to Model Context Protocol servers from any AI agent. The platform combines a public registry of thousands of community MCP servers with a managed gateway that handles OAuth, credential storage, session state, protocol compliance, and webhook triggers — so agents can call tools across servers via a single stateless Connect API. Smithery also distributes a CLI, a TypeScript API client, a local runner, a Skills Registry, deep-linking integrations for popular MCP clients (Claude Desktop, Cursor, Continue), and uplink support for exposing local servers as hosted connections.
2 APIs
1 Capabilities
27 Features
AIAgentsMCPModel Context ProtocolRegistryHostingToolsSkillsMarketplaceDeveloper Platform
Smithery publishes 2 APIs on the APIs.io network: Registry API and Connect API. Tagged areas include AI, Agents, MCP, Model Context Protocol, and Registry.
The Smithery catalog on APIs.io includes 1 machine-runnable capability, 1 JSON-LD context, and 1 Spectral governance ruleset.
Smithery’s developer surface includes developer portal, documentation, getting-started guide, code examples, CLI, tooling, SDKs, and 47 more developer resources.
Run Capabilities with Naftiko — Deploy and orchestrate these API capabilities using Naftiko Fleet.
Run with Naftiko
Run Capabilities with Naftiko — Deploy and orchestrate these API capabilities using Naftiko Fleet.
Run with Naftiko
Public MCP server registry with thousands of community-published MCP servers
Server publishing via URL (Streamable HTTP) or MCPB bundle for stdio-based servers
Managed gateway handling MCP protocol compliance, metadata enrichment, and caching
Auto-generated OAuth UI modals for servers requiring API keys or user configuration
Stateless Connect API — agents talk to MCP servers without holding session state client-side
Zero OAuth setup with encrypted credential storage and automatic credential refresh
Skills Registry — reusable prompt-based skills installable via `npx skills add`
Namespaces for grouping servers, connections, and skills under a user or organization
Triggers — expose MCP server events as webhooks; manage subscriptions per-connection or per-namespace
Uplink — expose a local MCP server as a hosted connection without deploying it
Deep linking for one-click client integration (Claude Desktop, Cursor, Continue, etc.)
Scoped service tokens for machine-to-machine access with per-namespace and per-connection scoping
Team API keys with admin-only create/list/revoke flow under organization namespaces
Custom domains for hosted MCP servers, managed at the registry level
Per-server secrets management for storing API keys and credentials used at runtime
Release management with logs, streaming SSE log feed, and resume-on-paused-release support
Runtime log retrieval for hosted MCP servers
MCP JSON-RPC endpoint exposed per connection at /connect/{namespace}/{connectionId}/mcp
Cross-server tool search at /tools — find any tool across the entire registry
Smithery CLI (`smithery`) installable via npm, Homebrew, or Scoop for managing MCP servers and skills
smithery runner — local MCP launcher for stdio servers
smithery-cli-mcp — official MCP server for the Smithery CLI itself
TypeScript API client (Apache-2.0)
Integration with Vercel AI SDK, Claude Desktop, Cursor, Continue, and any MCP-compatible client
agent.pw — share APIs with agents without sharing secrets
mouseless — Rust MCP server for macOS desktop control
hylo workflow engine with hylo-plugins for managed-agent and cloud-claude
aid: smithery-ai
url: https://raw.githubusercontent.com/api-evangelist/smithery-ai/refs/heads/main/apis.yml
apis:
- aid: smithery-ai:smithery-registry-api
name: Smithery Registry API
tags:
- MCP
- Registry
- Servers
- Skills
- Tools
- AI
- Agents
humanURL: https://smithery.ai/docs
baseURL: https://api.smithery.ai
properties:
- url: https://smithery.ai/docs
type: Documentation
- url: https://smithery.ai/docs/openapi.json
type: OpenAPI
- url: openapi/smithery-registry-api-openapi.yml
type: OpenAPI
- url: json-schema/smithery-server-schema.json
type: JSONSchema
- url: json-structure/smithery-server-structure.json
type: JSONStructure
- url: json-ld/smithery-ai-context.jsonld
type: JSONLD
- url: vocabulary/smithery-ai-vocabulary.yml
type: Vocabulary
- url: rules/smithery-ai-rules.yml
type: Spectral
description: Discover, publish, and manage MCP (Model Context Protocol) servers and Agent Skills on Smithery.
Browse the public registry of thousands of MCP servers, search tools across servers, publish your own
server releases (URL or MCPB bundle), manage release logs and runtime logs, set server secrets, manage
custom domains, create namespaces, manage organization API keys, and mint scoped service tokens for
machine-to-machine access.
- aid: smithery-ai:smithery-connect-api
name: Smithery Connect API
tags:
- MCP
- Connections
- Triggers
- Webhooks
- AI
- Agents
humanURL: https://smithery.ai/docs/usage/connect
baseURL: https://api.smithery.ai
properties:
- url: https://smithery.ai/docs/usage/connect
type: Documentation
- url: openapi/smithery-connect-api-openapi.yml
type: OpenAPI
- url: json-schema/smithery-connection-schema.json
type: JSONSchema
- url: json-structure/smithery-connection-structure.json
type: JSONStructure
- type: NaftikoCapability
url: capabilities/discover-and-connect.yaml
description: Connect AI agents to any MCP server hosted on Smithery without managing OAuth flows, credential
storage, or session state. Create stateless connections under a namespace, execute MCP JSON-RPC against
a Smithery-managed gateway, list and call tools, manage trigger definitions exposed by a server, and
subscribe (or unsubscribe) to trigger events that fire as webhooks. The Connect API is the runtime
counterpart to the Registry — it is how agents actually talk to the servers they discover.
name: Smithery
tags:
- AI
- Agents
- MCP
- Model Context Protocol
- Registry
- Hosting
- Tools
- Skills
- Marketplace
- Developer Platform
kind: contract
image: https://kinlane-productions2.s3.amazonaws.com/apis-json/apis-json-logo.jpg
access: 3rd-Party
common:
- url: https://smithery.ai
type: Portal
- url: https://smithery.ai/docs
name: Smithery Documentation
type: Documentation
- url: https://smithery.ai/docs/main
name: Introduction
type: GettingStarted
- url: https://smithery.ai/docs/openapi.json
name: Smithery Platform OpenAPI
type: OpenAPI
- url: https://smithery.ai/docs/llms.txt
name: Docs index (llms.txt)
type: Documentation
- url: https://smithery.ai/docs/mcp
name: Docs MCP Server
type: Documentation
- url: https://smithery.ai/docs/concepts/what_is_mcp
name: What is MCP
type: Documentation
- url: https://smithery.ai/docs/concepts/namespaces
name: Namespaces
type: Documentation
- url: https://smithery.ai/docs/concepts/cli
name: Smithery CLI
type: Documentation
- url: https://smithery.ai/docs/build
name: Publish MCP Servers
type: Documentation
- url: https://smithery.ai/docs/build/publish
name: Publishing Guide
type: Documentation
- url: https://smithery.ai/docs/build/triggers
name: Triggers
type: Documentation
- url: https://smithery.ai/docs/usage/connect
name: Connect API Usage
type: Documentation
- url: https://smithery.ai/docs/usage/deep-linking
name: Deep Linking
type: Documentation
- url: https://smithery.ai/docs/usage/uplink
name: Uplink (Local Server Exposure)
type: Documentation
- url: https://smithery.ai/docs/usage/token-scoping
name: Token Scoping
type: Documentation
- url: https://smithery.ai/docs/usage/listing_your_client
name: Listing Your Client
type: Documentation
- url: https://smithery.ai/docs/integrations/vercel_ai_sdk
name: Vercel AI SDK Integration
type: Documentation
- url: https://smithery.ai/docs/cookbooks/typescript_oauth_client
name: TypeScript OAuth Client Cookbook
type: CodeExamples
- url: https://github.com/smithery-ai
name: Smithery on GitHub
type: GitHubOrganization
- url: https://github.com/smithery-ai/cli
name: Smithery CLI
type: CLI
- url: https://github.com/smithery-ai/runner
name: smithery runner (local MCP launcher)
type: Tool
- url: https://github.com/smithery-ai/typescript-api
name: TypeScript API client
type: SDK
- url: https://github.com/smithery-ai/registry
name: Smithery Registry (public)
type: Tool
- url: https://github.com/smithery-ai/smithery-cli-mcp
name: Official MCP server for Smithery CLI
type: Tool
- url: https://github.com/smithery-ai/smithery-cookbook
name: Smithery Cookbook
type: CodeExamples
- url: https://github.com/smithery-ai/mcp-chat
name: mcp-chat (reference chat client)
type: Tool
- url: https://github.com/smithery-ai/agent.pw
name: agent.pw (share APIs with agents without sharing secrets)
type: Tool
- url: https://github.com/smithery-ai/mouseless
name: mouseless (Rust MCP server for macOS desktop control)
type: Tool
- url: https://github.com/smithery-ai/render-markdown
name: render-markdown
type: Tool
- url: https://github.com/smithery-ai/mcp-to-cli
name: mcp-to-cli (expose MCP servers as CLIs)
type: Tool
- url: https://github.com/smithery-ai/migration-guide-mcp
name: STDIO to Streamable HTTP migration MCP
type: Tool
- url: https://github.com/smithery-ai/agent-hook
name: agent-hook (install Claude Code hooks from GitHub)
type: Tool
- url: https://github.com/smithery-ai/homebrew-smithery
name: Homebrew tap
type: Documentation
- url: https://github.com/smithery-ai/scoop-smithery
name: Scoop bucket
type: Documentation
- url: https://github.com/smithery-ai/skills
name: Smithery skills
type: Tool
- url: https://github.com/smithery-ai/hylo
name: hylo (workflow engine)
type: Tool
- url: https://github.com/smithery-ai/hylo-plugins
name: hylo plugins (managed-agent, cloud-claude)
type: Plugins
- url: https://github.com/smithery-ai/mouseless
name: mouseless
type: Tool
- url: https://github.com/smithery-ai/familysearch-mcp
name: FamilySearch MCP server
type: Tool
- url: https://github.com/smithery-ai/mcp-server-7
name: DevRev MCP server
type: Tool
- url: https://github.com/smithery-ai/mcp-multilspy
name: mcp-multilspy
type: Tool
- url: https://github.com/smithery-ai/mcp-oauth-debug
name: mcp-oauth-debug
type: Tool
- url: https://github.com/smithery-ai/openchat
name: openchat
type: Tool
- url: https://github.com/smithery-ai/okay-error
name: okay-error
type: Tool
- url: https://github.com/smithery-ai/workers-biscuit
name: Biscuit tokens for Cloudflare Workers
type: Tool
- url: https://github.com/smithery-ai/rfm
name: Request For MCPs
type: Documentation
- url: https://github.com/modelcontextprotocol
name: Model Context Protocol
type: Documentation
- url: https://modelcontextprotocol.io
name: MCP specification
type: Documentation
- url: https://smithery.ai/pricing
name: Pricing
type: Pricing
- url: https://plans/smithery-ai-plans-pricing.yml
type: Plans
- url: https://rate-limits/smithery-ai-rate-limits.yml
type: RateLimits
- url: https://finops/smithery-ai-finops.yml
type: FinOps
- type: NaftikoCapability
url: capabilities/discover-and-connect.yaml
- type: Features
data:
- Public MCP server registry with thousands of community-published MCP servers
- Server publishing via URL (Streamable HTTP) or MCPB bundle for stdio-based servers
- Managed gateway handling MCP protocol compliance, metadata enrichment, and caching
- Auto-generated OAuth UI modals for servers requiring API keys or user configuration
- Stateless Connect API — agents talk to MCP servers without holding session state client-side
- Zero OAuth setup with encrypted credential storage and automatic credential refresh
- Skills Registry — reusable prompt-based skills installable via `npx skills add`
- Namespaces for grouping servers, connections, and skills under a user or organization
- Triggers — expose MCP server events as webhooks; manage subscriptions per-connection or per-namespace
- Uplink — expose a local MCP server as a hosted connection without deploying it
- Deep linking for one-click client integration (Claude Desktop, Cursor, Continue, etc.)
- Scoped service tokens for machine-to-machine access with per-namespace and per-connection scoping
- Team API keys with admin-only create/list/revoke flow under organization namespaces
- Custom domains for hosted MCP servers, managed at the registry level
- Per-server secrets management for storing API keys and credentials used at runtime
- Release management with logs, streaming SSE log feed, and resume-on-paused-release support
- Runtime log retrieval for hosted MCP servers
- MCP JSON-RPC endpoint exposed per connection at /connect/{namespace}/{connectionId}/mcp
- Cross-server tool search at /tools — find any tool across the entire registry
- Smithery CLI (`smithery`) installable via npm, Homebrew, or Scoop for managing MCP servers and skills
- smithery runner — local MCP launcher for stdio servers
- smithery-cli-mcp — official MCP server for the Smithery CLI itself
- TypeScript API client (Apache-2.0)
- Integration with Vercel AI SDK, Claude Desktop, Cursor, Continue, and any MCP-compatible client
- agent.pw — share APIs with agents without sharing secrets
- mouseless — Rust MCP server for macOS desktop control
- hylo workflow engine with hylo-plugins for managed-agent and cloud-claude
sources:
- https://smithery.ai/docs
- https://smithery.ai/docs/openapi.json
- https://smithery.ai/docs/llms.txt
- https://github.com/smithery-ai
- https://github.com/smithery-ai/cli
updated: '2026-05-25'
created: '2026-05-25'
modified: '2026-05-25'
position: Producing
description: Smithery is an MCP server registry and hosting platform that lets developers discover, publish,
and connect to Model Context Protocol servers from any AI agent. The platform combines a public registry of
thousands of community MCP servers with a managed gateway that handles OAuth, credential storage, session
state, protocol compliance, and webhook triggers — so agents can call tools across servers via a single
stateless Connect API. Smithery also distributes a CLI, a TypeScript API client, a local runner, a Skills
Registry, deep-linking integrations for popular MCP clients (Claude Desktop, Cursor, Continue), and uplink
support for exposing local servers as hosted connections.
maintainers:
- FN: Kin Lane
email: [email protected]
X: apievangelist
url: https://apievangelist.com
specificationVersion: '0.16'