Client compatibility

This page shows which AI clients work with ToolHive and how they connect to MCP servers.

Supported clients

We've tested ToolHive with these clients:

Client	Supported	Auto-configuration	Notes
GitHub Copilot (VS Code)	✅	✅	v1.100+ or Insiders version (v1.102+, see note)
Cursor	✅	✅	v0.50.0+
Roo Code (VS Code)	✅	✅	v3.19.2+
Cline (VS Code)	✅	✅	v3.8.5+ (sse only; streamable-http issue)
Claude Code CLI	✅	✅	v1.0.27+
GitHub Copilot (JetBrains)	✅	❌	v1.5.47+
Continue (VS Code)	✅	❌	v1.0.14+
Continue (JetBrains)	✅	❌	v1.0.23+
PydanticAI	✅	❌	v0.2.18+
ChatGPT Desktop	❌	❌	No support for HTTP/SSE MCPs
Claude Desktop	❌	❌	No support for HTTP/SSE MCPs (issue)

The minimum versions listed are the earliest versions that support the Streamable HTTP transport protocol.

You can also use other clients and development libraries that support the SSE and/or Streamable HTTP protocol, but you'll need to configure them manually.

Client requirements

To work with ToolHive, your client needs to:

1. Support the Model Context Protocol (MCP)
2. Connect to MCP servers using the server-sent events (SSE) or Streamable HTTP transport protocol
3. Have the correct MCP server URL configured

Automatic configuration support

ToolHive can automatically configure supported clients to connect to MCP servers. See the UI client configuration or CLI client configuration guide for more details.

Check the table above to see which clients support automatic configuration.

For other clients, you'll need to set up the MCP server URL yourself.

Configuration locations

ToolHive manages client configurations in their standard locations.

VS Code with Copilot

GitHub Copilot in VS Code stores its global MCP configuration in your VS Code user settings file.

note

VS Code version 1.102 moved the MCP settings to a new location. Servers are now configured in a dedicated mcp.json file in the user settings directory.

ToolHive currently only supports the previous location (settings.json). When ToolHive adds a new MCP server, VS Code prompts you to move it to the new location.

This works, but ToolHive will not automatically update the new mcp.json file in the same location as the settings.json file referenced below. You can still use the MCP servers, but you will need to manually update the mcp.json file if you stop or remove the server.

Subscribe to this issue for updates on ToolHive's support for the new MCP configuration location.

Standard version:

• macOS: ~/Library/Application Support/Code/User/settings.json
• Linux: ~/.config/Code/User/settings.json

Insiders edition:

• macOS: ~/Library/Application Support/Code - Insiders/User/settings.json
• Linux: ~/.config/Code - Insiders/User/settings.json

Example configuration:

{
  // Other VS Code settings...

  "mcp": {
    "servers": {
      "github": { "url": "http://localhost:19046/sse#github", "type": "sse" },
      "fetch": { "url": "http://localhost:43832/sse#fetch", "type": "sse" },
      "osv": { "url": "http://localhost:51712/mcp", "type": "http" }
    }
  }
}

When you register VS Code as a client, ToolHive automatically updates the global MCP configuration file whenever you run an MCP server. You can also configure project-specific MCP servers by creating a .vscode/mcp.json file in your project directory.

Cursor

Cursor stores its global MCP configuration in a JSON file in your home directory.

• All platforms: ~/.cursor/mcp.json

Example configuration:

{
  "mcpServers": {
    "github": { "url": "http://localhost:19046/sse#github" },
    "fetch": { "url": "http://localhost:43832/sse#fetch" },
    "osv": { "url": "http://localhost:51712/mcp" }
  }
}

Cursor automatically detects the transport mode for HTTP MCP servers.

When you register Cursor as a client, ToolHive automatically updates the global MCP configuration file whenever you run an MCP server. You can also configure project-specific MCP servers by creating a .cursor/mcp.json file in your project directory.

Claude Code

Claude Code stores its global (user scope) MCP configuration in a JSON file in your home directory.

• All platforms: ~/.claude.json

Example configuration:

{
  // Other Claude Code settings...

  "mcpServers": {
    "github": { "url": "http://localhost:19046/sse#github", "type": "sse" },
    "fetch": { "url": "http://localhost:43832/sse#fetch", "type": "sse" },
    "osv": { "url": "http://localhost:51712/mcp", "type": "http" }
  }
}

When you register Claude Code as a client, ToolHive automatically updates the global MCP configuration file whenever you run an MCP server. You can also configure project-specific MCP servers by creating a .mcp.json file in your project directory, or add servers using the claude CLI:

claude mcp add --scope <user|project> --transport sse fetch http://localhost:43832/sse#fetch

Roo Code and Cline

Roo Code (previously Roo Cline) and Cline store their global MCP configuration in their VS Code extension settings directory. Both use the same configuration format.

Roo Code:

• macOS: ~/Library/Application Support/Code/User/globalStorage/rooveterinaryinc.roo-cline/settings/mcp_settings.json
• Linux: ~/.config/Code/User/globalStorage/rooveterinaryinc.roo-cline/settings/mcp_settings.json

Cline:

• macOS: ~/Library/Application Support/Code/User/globalStorage/saoudrizwan.claude-dev/settings/cline_mcp_settings.json
• Linux: ~/.config/Code/User/globalStorage/saoudrizwan.claude-dev/settings/cline_mcp_settings.json

Example configuration:

{
  "mcpServers": {
    "github": { "url": "http://localhost:19046/sse#github", "type": "sse" },
    "fetch": { "url": "http://localhost:43832/sse#fetch", "type": "sse" },
    "osv": { "url": "http://localhost:51712/mcp", "type": "streamable-http" }
  }
}

When you register Roo Code or Cline as a client, ToolHive automatically updates the global MCP configuration file whenever you run an MCP server. With Roo Code, you can also configure project-specific MCP servers by creating a .roo/mcp.json file in your project directory.

Manual configuration

If your client doesn't support automatic configuration, you'll need to set up the MCP server URL manually.

Example: PydanticAI

For PydanticAI, set the MCP server URL in your code:

from pydantic_ai.mcp import MCPServerSSE
from pydantic_ai.mcp import MCPServerStreamableHTTP

# For Server-Sent Events (SSE)
server = MCPServerSSE(url='http://localhost:43832/sse#fetch')

# For Streamable HTTP
server = MCPServerStreamableHTTP(url='http://localhost:51712/mcp')

Example: Continue

For the Continue extension in VS Code or JetBrains, edit your ~/.continue/config.yaml file or project-specific .continue/mcpServers/<name>.yaml file to include the MCP server URL:

mcpServers:
  - name: fetch
    type: sse
    url: http://localhost:43832/sse#fetch

Continue supports SSE (type: sse) and Streamable HTTP (type: streamable-http) transports.

Example: Copilot for JetBrains IDEs

For the GitHub Copilot plugin in JetBrains IDEs (IntelliJ, Pydantic, etc.), edit your ~/.config/github-copilot/mcp.json file to include the MCP server URL:

{
  "servers": {
    "fetch": { "url": "http://localhost:43832/sse#fetch", "type": "sse" },
    "osv": { "url": "http://localhost:51712/mcp", "type": "http" }
  }
}

Copilot for JetBrains supports SSE ("type": "sse") and Streamable HTTP ("type": "streamable-http") transports.

Related information

• Client configuration
• Run MCP servers