Install DocSearch MCP

DocSearch MCP is a remote MCP server. Point any MCP-compatible client at this endpoint — no authentication required:

https://mcp.algolia.com/1/docsearch/mcp

The fastest path to getting started is the DocSearch CLI — one command that configures your client for you. Prefer to set things up yourself? Install it as a plugin (ships the MCP server plus client guidance like rules, skills, and commands) or manually (just the MCP server config). Pick your client below.

Connect DocSearch MCP

Pick your client to get the install steps

One-command installAdds the server, rules, and skills to Cursor

$ npx @docsearch/cli setup --cursor

CursorPlugin · 1-click · manual

Install the Cursor plugin package from the DocSearch repository:

Plugin package

mcp/plugins/docsearch/cursor/algolia-docsearch

It ships the MCP server, a rule, and a skill that teach Cursor when and how to call the DocSearch tools.

Install a client plugin

Plugins include the MCP connection and instructions that teach the client when and how to use DocSearch.

Claude Code

Run these commands inside Claude Code:

/plugin marketplace add algolia/docsearch
/plugin install algolia-docsearch@algolia-docsearch-marketplace

The plugin adds the MCP server, a skill, and the /algolia-docsearch:docs command.

Cursor

The Cursor plugin package is in the DocSearch repository at:

mcp/plugins/docsearch/cursor/algolia-docsearch

It includes the MCP server configuration, a rule, and a skill.

Configure a supported agent manually

Manual configuration adds the server but doesn't install the rules and skills supplied by the CLI or plugins.

Cursor

Add this entry to ~/.cursor/mcp.json for user-level configuration or .cursor/mcp.json for project configuration:

.cursor/mcp.json

{
  "mcpServers": {
    "algolia-docsearch": {
      "url": "https://mcp.algolia.com/1/docsearch/mcp"
    }
  }
}

Claude Code

Add the user-level remote server with the Claude Code CLI:

claude mcp add --scope user --transport http algolia-docsearch https://mcp.algolia.com/1/docsearch/mcp

Codex

Add the remote server with the Codex CLI:

codex mcp add algolia-docsearch --url https://mcp.algolia.com/1/docsearch/mcp

You can also add the URL under [mcp_servers.algolia-docsearch] in ~/.codex/config.toml.

OpenCode

Add this entry to opencode.json or opencode.jsonc:

opencode.json

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "algolia-docsearch": {
      "type": "remote",
      "url": "https://mcp.algolia.com/1/docsearch/mcp",
      "enabled": true
    }
  }
}

Gemini CLI

Add this entry to .gemini/settings.json for project configuration or ~/.gemini/settings.json for user-level configuration:

.gemini/settings.json

{
  "mcpServers": {
    "algolia-docsearch": {
      "httpUrl": "https://mcp.algolia.com/1/docsearch/mcp"
    }
  }
}

Connect another MCP client

You can also connect clients that support remote HTTP MCP servers. Use these connection methods:

• Claude Desktop: Add a custom connector named Algolia DocSearch with the endpoint URL.
• ChatGPT: Add a no-auth custom connector. Availability depends on the plan and workspace settings.
• Pi: Install an MCP extension or adapter before adding the server.
• VS Code: Add a remote server to the workspace or user MCP configuration. The tools are available in Copilot Chat agent mode.
• Windsurf: Add the endpoint to the Cascade MCP configuration, then restart Windsurf.
• Zed: Add the endpoint as a context server.
• Conductor: Configure the server in the selected Claude Code, Codex, or Cursor host. Conductor doesn't use a separate MCP configuration for those sessions.
• Antigravity: Add the endpoint through Manage MCP Servers > View raw config, then refresh the installed servers.

Configuration keys differ by client. Follow the client's MCP documentation and keep the server name algolia-docsearch and the endpoint URL unchanged.

Verify the installation

Ask your client to search public documentation:

Use DocSearch MCP to find the Next.js middleware matcher documentation.

Confirm that the client calls a tool whose name starts with algolia_docsearch_ and returns source links from the selected documentation.

If verification fails, see Troubleshooting.