parallel-web/search-mcp

Parallel Search MCP

Free web search for your coding agent.

Find current documentation, check facts, and read web pages without leaving your agent. No account or API key needed to get started.

Ask your agent to install

Copy this into your coding agent:

Install Parallel's free Search MCP in the coding agent I'm using now.
Identify your client and use its native MCP setup instructions:
https://docs.parallel.ai/integrations/mcp/search-mcp

Server name: parallel-search
Server URL: https://search.parallel.ai/mcp
Transport: Streamable HTTP
Authentication: none. No account, API key, or Parallel CLI is needed.

Preserve existing settings and reuse any connection to this endpoint.
Tell me if a restart or approval is required.

Once connected, check that web_search and web_fetch are available.
Use Parallel to find the official FastAPI response-model documentation
and return a source link. Report anything you couldn't verify.

Install in Cursor · Install in VS Code · Manual setup · Documentation

Manual setup

Choose your client. You only need one of these options.

Claude Code

Run this in the project where you want to use Parallel:

claude mcp add --transport http --scope project parallel-search https://search.parallel.ai/mcp

Start a new Claude Code session, approve the MCP connection if prompted, and run /mcp to check that parallel-search is connected. The free endpoint does not require a Parallel login.

Codex

codex mcp add parallel-search --url https://search.parallel.ai/mcp

Start a new Codex session and run /mcp to check the connection. See the Codex MCP guide for configuration options.

Gemini CLI

gemini extensions install https://github.com/parallel-web/search-mcp

Restart Gemini CLI and run /mcp to check that parallel-search is connected. The extension uses Streamable HTTP and requires no Parallel account or API key. If you already configured parallel-search manually, that configuration takes precedence over the extension. See the extension guide for installation and update options.

OpenCode

Merge this into your project's opencode.json:

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "parallel-search": {
      "type": "remote",
      "url": "https://search.parallel.ai/mcp",
      "oauth": false
    }
  }
}

Restart OpenCode and run opencode mcp list to check the connection. See the OpenCode MCP guide for details.

OpenClaw

Follow the OpenClaw setup guide to install the Parallel plugin, then set tools.web.search.provider to parallel-free. No Parallel account or API key is needed.

Cursor, VS Code, and other clients

Use this remote MCP server URL:

https://search.parallel.ai/mcp

Select HTTP or Streamable HTTP if your client asks for a transport. Follow the client-specific instructions for the right settings file and format.

Try it

Ask your agent:

Use Parallel to find the official FastAPI docs for response models. Show me a small example and link to the source.

Or try:

Use Parallel to check the latest stable Node.js release and its release notes. Cite the official sources.

Use Parallel to read https://modelcontextprotocol.io/introduction and explain how MCP connects an agent to tools.

Your agent can use two tools:

Tool	What it does
web_search	Searches the web and returns relevant excerpts with source links.
web_fetch	Reads specific URLs and returns page content for your agent to use.

Search results often contain enough information to answer. Your agent can fetch a page when it needs more detail.

Free access and higher limits

This plugin is anonymous by design: both tools work with no account, API key, or login, subject to free-tier rate limits. The free tier is intended for personal agents, hobby projects, and exploration.

For higher limits, usage analytics, and production workloads, create an account at platform.parallel.ai and connect the Parallel Search connector, which signs in with that account. Authenticated usage is billed under your account's pricing and limits — see parallel.ai for pricing. Don't add an API key or point this plugin at an authenticated endpoint; the connector handles sign-in natively.

Both tiers include the same two tools. Authenticated connections also let you configure search behavior, including search modes, result counts, and domain filters. Anonymous requests ignore these settings.

Privacy and terms

Parallel hosts the service. Search queries and requested URLs are sent to Parallel and handled under the Customer Terms and Privacy Policy. The server only receives what the tools send it — search queries and the URLs you ask it to fetch — and does not read your conversation, files, or other context.

Troubleshooting

• Tools aren't showing up: start a new agent session and check its MCP connection list. Confirm the URL is https://search.parallel.ai/mcp.
• Asked to log in to Parallel: check the URL. /mcp supports free access; /mcp-oauth requires authentication. Approving an MCP connection in your client is separate from logging in to Parallel.
• Hit a rate limit: wait before retrying, or move to an authenticated Parallel account via the Parallel Search connector for higher limits.
• Still stuck: check the setup guide or open an issue. Include your client, version, and error message. Leave out API keys and other credentials.

Support and security

• Questions and bugs: open an issue with your client, version, and error message (leave out API keys and other credentials), or see the documentation.
• Product support: contact Parallel at [email protected].
• Security concerns: report privately to [email protected] — see SECURITY.md. Please don't file security reports as public issues.

About this repo

This is the official GitHub home for Parallel's hosted Search MCP. It contains this guide and Claude, Codex, and Cursor plugin configurations. All three use the same .mcp.json. The Claude plugin also includes a setup skill (skills/parallel-search-setup/SKILL.md) that walks Claude through verifying the connection, using the tools well, troubleshooting, and moving to an authenticated Parallel account when higher limits are needed. You can suggest documentation fixes and examples here. The search service runs on Parallel's infrastructure; cloning this repo gives you the documentation and configuration files.

For CLI-based search, research, and enrichment, see the separate Parallel CLI skills plugin.

To try the Claude plugin from a local checkout:

git clone https://github.com/parallel-web/search-mcp.git
claude --plugin-dir ./search-mcp

Use either the plugin or the direct MCP setup above to avoid adding the same server twice.

The files in this repo are MIT licensed. Hosted service usage is covered by the Customer Terms linked above.