Cline MCP web search guide
How to Add Web Search to Cline with Webstractor MCP
Cline can inspect a repository and edit code, but a search MCP gives it an explicit way to discover current external sources before making a technical decision. Webstractor is especially useful for finding official documentation, recent migration notes, and public examples while keeping the original URLs in view.
Quick answer
Add a hosted server from Cline’s Remote Servers tab.
Open MCP Servers in Cline, choose the Remote Servers tab, enter https://webstractor.com/mcp, and select Streamable HTTP. In JSON, Cline spells the transport streamableHttp. Keep autoApprove empty until you have reviewed the read-only search tools.
{
"mcpServers": {
"webstractor": {
"type": "streamableHttp",
"url": "https://webstractor.com/mcp",
"disabled": false,
"autoApprove": []
}
}
}Before you begin
What you need
- Cline for VS Code, JetBrains, or the Cline CLI
- A configured model/provider
- A workspace where you can inspect MCP tool calls
Step-by-step setup
Connect Webstractor to Cline
Open the MCP Servers screen
Click the stacked-server icon in Cline. For the IDE extension, the Remote Servers tab is the least error-prone path; CLI users can run cline mcp.
Select Streamable HTTP
Enter Webstractor as the unique server name and the full /mcp endpoint as the URL. Do not accept the legacy SSE default.
Use the exact JSON spelling when configuring manually
Cline uses streamableHttp in its JSON, unlike several other MCP clients. Its IDE config or ~/.cline/mcp.json can contain this entry.
{
"mcpServers": {
"webstractor": {
"type": "streamableHttp",
"url": "https://webstractor.com/mcp",
"disabled": false,
"autoApprove": []
}
}
}Confirm search_web appears
Cline should list the tool after connecting. Keep autoApprove empty during setup so the first proposed query is visible before execution.
Validate with a code-research prompt
Prompt: “Before editing anything, use search_web with site=docs.astro.build to find current Astro middleware documentation. Return the five most relevant URLs and explain which one answers the question.”
Safe workflow
Separate research from code changes
Ask Cline to search and prepare a source-backed plan before it edits files. This creates a useful checkpoint: you can inspect the sources, reject irrelevant results, and only then authorize implementation.
A site filter is valuable when package names are ambiguous or third-party tutorials lag behind the official docs.
Tool policy
Start with explicit approval
Cline supports per-server autoApprove lists, but an empty list is the best default for a new integration. Search is read-only, yet reviewing arguments still catches accidental secrets and overly broad queries.
- Inspect the query string
- Prefer official domains
- Limit the result count
- Approve edits separately from research
Troubleshooting
Cline’s transport name is case-sensitive
If a manually added server silently falls back or fails, confirm type is exactly streamableHttp. Omitting it defaults to legacy SSE for backward compatibility.
If tools load but Cline uses its browser instead, name search_web directly and ask it not to edit until the research is complete.
Available data
What you can extract
- Ranked source URLs and concise snippets
- Site-restricted searches for official documentation
- Language and country controls
- A direct path from discovery to full-page extraction
AI workflows
Where normalized data helps
- Research a dependency before changing code
- Find an official migration guide for a failing upgrade
- Locate current browser or runtime compatibility notes
- Collect candidate sources for an implementation plan
Boundaries
Public data only
- Search results do not authorize Cline to change project files.
- Leave MCP auto-approval off unless your risk model permits it.
- Public search queries should not include repository secrets or customer data.
webstractor.com does not bypass CAPTCHAs, login walls, paywalls, access controls, or regional restrictions. Review the source’s terms and applicable law before collecting or reusing data.
Common questions
Cline and Webstractor FAQ
Where is Cline’s MCP config?
The Cline CLI uses ~/.cline/mcp.json. In the IDE, open MCP Servers → Configure → Configure MCP Servers, or use the Remote Servers tab.
Can I commit the server configuration?
Keep project policy in version control only if it contains no credentials. Never commit an account API key in headers.
Does Webstractor need an API key?
Anonymous allowance works without one. Account usage can add an Authorization: Bearer ext_live_… header in clients that support remote headers.
Ready to try it?