Kilo Code URL extraction / 2026
How to Fetch Web Pages in Kilo Code with Webstractor MCP in 2026
Kilo Code can edit a repository, but a linked web page is outside that workspace. A dedicated extraction tool gives it clean public-page context while keeping the operation read-only and visible in the Permission Dock. Use it for a URL already named in an issue, TODO, dependency note, or conversation.
Quick answer
Attach one read-only MCP tool to Kilo’s agent.
Add the hosted endpoint as a remote server, confirm extract_public_url is available, and give Kilo the exact page URL. Request Markdown plus a focused topic when it needs a specific part of long documentation.
{
"mcp": {
"webstractor": {
"type": "remote",
"url": "https://webstractor.com/mcp",
"enabled": true
}
}
}Before you begin
What you need
- Kilo Code extension or CLI
- A public page URL
- Permission to add a project or global MCP server
Step-by-step setup
Connect Webstractor to Kilo Code
Connect the hosted server
Open Settings → MCP → Add Server → Remote (HTTP). Use `webstractor` as the name and the hosted /mcp URL. No command, environment variable, or local package is required.
{
"name": "webstractor",
"type": "remote",
"url": "https://webstractor.com/mcp"
}Confirm the extraction tool is enabled
Inspect the server in Settings → Agent Behaviour → MCP Servers. The connection must be active and extract_public_url must not be denied by a permission rule.
Test one precise read
Supply an exact HTTPS URL and request attribution. Approve the tool call after checking the URL and focus arguments in the Permission Dock.
# Paste into Kilo Code
Use webstractor extract_public_url to read https://kilo.ai/docs/automate/mcp/using-in-kilo-code as Markdown with focus "remote HTTP configuration". Produce a five-step checklist and cite the source URL.Safe automation
Auto-approve the reader, not every remote tool
If the page reader becomes part of a repeated workflow, Kilo can allow `webstractor_extract_public_url` while leaving other tools confirm-first. This retains control over the exact URL that leaves the editor.
Do not put secrets into a URL. Query strings can be logged by clients and services; the extraction tool is intended for ordinary public pages.
{
"permission": {
"webstractor_extract_public_url": "allow"
}
}Troubleshooting
Test the connection and the URL separately
If every call fails, inspect the MCP server indicator and confirm the remote type and endpoint. If only one page fails, try a stable public documentation URL; the first source may be unavailable or restricted.
When the output is too short, remove the focus hint. When it is too broad, use a concise topic such as “configuration precedence” rather than a full question.
Available data
What you can extract
- Clean page Markdown
- Optional topic-focused content
- Normalized schema-v1 JSON entities
- Canonical URLs and source metadata
AI workflows
Where normalized data helps
- Read an issue link before reproducing a bug
- Extract a migration checklist from official docs
- Summarize a public technical article beside code
- Review current terms or product documentation
Boundaries
Public data only
- Extraction reads one URL and does not follow a site recursively.
- Authenticated or restricted pages remain unavailable.
- A source can change; keep the canonical URL beside conclusions.
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
Kilo Code and Webstractor FAQ
Can Kilo Code fetch a link from an issue?
Yes, if it is a publicly reachable HTTP or HTTPS URL. Ask Kilo to pass the exact link to extract_public_url.
Will it crawl every linked page?
No. The tool extracts one supplied URL per call.
Can I require approval for every fetch?
Yes. Leave the default permission at confirm and inspect each URL in Kilo’s Permission Dock.
Ready to try it?