This repository contains the registry of MCP (Model Context Protocol) servers available for ToolHive. Each server entry provides AI assistants with specialized tools and capabilities.

Think of this as a catalog of tools that AI assistants can use. Each entry in this registry represents a server that provides specific capabilities—like interacting with GitHub, querying databases, or fetching web content.

Adding your MCP server to the registry is simple! You just need to create a YAML file with some basic information about your server. We support two types of MCP servers:

1. Container-based servers - Run as Docker containers
2. Remote servers - Accessed via HTTP/HTTPS endpoints

Create a new folder in the registry/ directory with your server's name (use lowercase and hyphens):

registry/ └── my-awesome-server/ └── spec.yaml 

Choose the appropriate format based on your server type:

Create a spec.yaml file with this minimum information:

# Required fields - you must provide these image: docker.io/myorg/my-server:latest # Your Docker image description: What your server does in one sentence transport: stdio # How your server communicates (usually "stdio") tier: Community # "Official" or "Community" status: Active # "Active" or "Deprecated" tools: - tool_name_1 # List the tools your server provides - tool_name_2 # Or use "set_during_runtime" if tools aren't knowable up-front # Recommended fields - helps users understand your server repository_url: https://github.com/myorg/my-server # Where to find your code

Create a spec.yaml file with this minimum information:

# Required fields - you must provide these url: https://api.example.com/mcp # Your MCP server endpoint description: What your server does in one sentence transport: streamable-http # Remote servers use "streamable-http" (preferred) or "sse" (not "stdio") tier: Community # "Official" or "Community" status: Active # "Active" or "Deprecated" tools: - tool_name_1 # List the tools your server provides - tool_name_2 # Or use "set_during_runtime" if tools aren't knowable up-front # Recommended fields - helps users understand your server repository_url: https://github.com/myorg/my-server # Where to find your code

You can add more information to help users:

# ... required fields above ... # Tell users what environment variables they need env_vars: - name: API_KEY description: Your API key from example.com required: true secret: true # Mark sensitive data - name: TIMEOUT description: Request timeout in seconds required: false default: "30" # Help users find your server tags: - api - integration - productivity # Server classification tier: Community # or "Official" if maintained by the protocol team status: Active # or "Deprecated"

image: ghcr.io/github/github-mcp-server:v0.10.0 description: Provides integration with GitHub's APIs for repository management transport: stdio repository_url: https://github.com/github/github-mcp-server tools: - create_issue - create_pull_request - get_file_contents - search_repositories env_vars: - name: GITHUB_PERSONAL_ACCESS_TOKEN description: GitHub personal access token with appropriate permissions required: true secret: true tags: - github - version-control - api tier: Official status: Active

url: https://api.example.com/mcp/v1 description: Provides access to Example API services via MCP transport: streamable-http repository_url: https://github.com/example/mcp-server tools: - fetch_data - process_request - submit_job # Authentication headers for remote servers headers: - name: X-API-Key description: API key for authentication required: true secret: true # OAuth configuration (alternative to headers) oauth_config: issuer: https://auth.example.com client_id: mcp-client scopes: - read - write tags: - api - remote - cloud tier: Community status: Active

This tells ToolHive how to communicate with your server:

For container-based servers:

• stdio - Standard input/output (most common)
• sse - Server-sent events
• streamable-http - HTTP streaming

For remote servers:

• streamable-http - HTTP streaming (recommended)
• sse - Server-sent events (deprecated but still supported)
• Note: Remote servers cannot use stdio

If you're not sure, use stdio for containers and streamable-http for remote servers.

Important: For container servers using streamable-http or sse transport, you must also specify the target_port field.

• Official - Maintained by the MCP team or platform owners
• Community - Created and maintained by the community (most servers)

• Active - Fully functional and maintained
• Deprecated - No longer maintained, will be removed

List all the tools your server provides. If your server's tools aren't knowable until runtime (for example, if they're dynamically generated based on configuration), you can use "set_during_runtime" as the value instead of listing specific tool names.

For container-based servers: Yes. Your MCP server must be packaged as a Docker image and published to a registry like:

• Docker Hub (docker.io/username/image)
• GitHub Container Registry (ghcr.io/username/image)
• Other public registries

For remote servers: No. You just need to provide the URL endpoint where your MCP server is accessible.

Important: Don't specify filesystem paths or volume mounts in your registry entries. Mounting host directories is a security risk and should be configured by users at runtime, not in registry specs. Only network permissions (allowed hosts and ports) should be specified in the permissions section.

After adding your entry, you can validate it:

# If you have the build tools installed task validate

Or submit a pull request and our automated checks will validate it for you.

1. Fork this repository
2. Add your server entry as described above
3. Submit a pull request
4. We'll review and merge your addition

We evaluate submissions based on several criteria to ensure quality and usefulness for the community. Your server should:

• Provide clear value and unique capabilities
• Be well-documented with accurate tool descriptions
• Follow security best practices
• Be actively maintained with recent updates

For detailed evaluation criteria, see the Registry Criteria documentation.

For container-based servers:

• Your Docker image is publicly accessible
• The description clearly explains what your server does
• You've listed all the tools your server provides
• Any required environment variables are documented
• Your server works with ToolHive

For remote servers:

• Your server endpoint is publicly accessible
• The description clearly explains what your server does
• You've listed all the tools your server provides
• Any required authentication (headers/OAuth) is documented
• The transport is set to streamable-http (preferred) or sse (not stdio)
• Your server works with ToolHive's proxy command

• Check existing entries in the registry/ folder for examples
• Open an issue if you have questions
• Join our community discussions

If you need to work with the registry programmatically:

# Import existing registry task import # Validate all entries task validate # Build the registry.json task build:registry # See all available commands task

Apache License 2.0