BoltAI URL extraction / 2026
How to Fetch Web Pages in BoltAI with Webstractor MCP in 2026
BoltAI is often used with selected text from other Mac apps, but a current public URL needs an explicit retrieval step. A Webstractor connector lets a dedicated agent read the page cleanly, without inheriting browser cookies or enabling interactive browser automation. Per-tool selection in BoltAI 2.x keeps the integration limited to the page reader.
Quick answer
Add the connector, then enable only the page reader.
Connect the hosted MCP endpoint in BoltAI’s Connectors settings and enable extract_public_url for your chosen agent. Pass the exact public page, request Markdown, and ask BoltAI to keep the source URL beside its summary.
{
"mcpServers": {
"webstractor": {
"command": "npx",
"args": ["-y", "mcp-remote", "https://webstractor.com/mcp"]
}
}
}Before you begin
What you need
- BoltAI 2.x recommended
- A tool-capable model
- Node.js only for the mcp.json fallback
Step-by-step setup
Connect Webstractor to BoltAI
Add Webstractor as a remote connector
Open Settings → Connectors and add the hosted /mcp endpoint as a custom server. Name it webstractor. If your BoltAI version only provides Plugins, use the compatibility configuration below.
Configure the plugin fallback if necessary
Open Settings → Plugins, click plus, paste the entry, save, then Reload Config. This starts a local transport bridge whenever BoltAI activates the server.
{
"mcpServers": {
"webstractor": {
"command": "npx",
"args": ["-y", "mcp-remote", "https://webstractor.com/mcp"]
}
}
}Enable the reader and test a public URL
Select only extract_public_url in the agent’s tools when that is all the profile needs. The test asks for focused Markdown and an attributable result.
# Paste into BoltAI
Use webstractor extract_public_url to read https://boltai.com/docs/plugins/mcp-servers as Markdown with focus "remote MCP servers". Summarize the current and legacy setup paths and cite the URL.Workflow design
Use selected text and fetched pages for different jobs
BoltAI’s system-wide selection tools are ideal when the relevant text is already visible in another app. Use Webstractor when the assistant needs to read the current public URL itself, especially when a full heading structure or canonical source matters.
For a reusable writing agent, instruct it to separate extracted facts from suggestions and always include the URL in final notes.
- Visible selected text → BoltAI selection workflow
- Known public link → extract_public_url
- Unknown source → search_web, then extract
Troubleshooting
Confirm the tool is enabled for this agent
BoltAI can install a server without exposing every tool to every agent. Check the active agent’s plugin dropdown and select extract_public_url.
For the legacy bridge, run the npx command in Terminal if it fails silently. If only one source fails, test a stable public page before changing MCP configuration.
Available data
What you can extract
- Readable Markdown from a known page
- Focused content for a named topic
- Semantic JSON for structured workflows
- Canonical URL and available public metadata
AI workflows
Where normalized data helps
- Summarize a linked document before writing
- Read current product documentation
- Compare public pages from two vendors
- Turn a long article into source-linked notes
Boundaries
Public data only
- The connector reads public pages, not content behind a login.
- One call extracts one URL and does not crawl a site.
- Local MCP configuration is device-specific and may not sync.
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
BoltAI and Webstractor FAQ
Can BoltAI read a public URL without browser access?
Yes. extract_public_url retrieves the public page through the hosted MCP server; it does not use your browser session.
Which output should I choose?
Use Markdown for summaries and reasoning. Use JSON for a workflow that consumes normalized fields.
Will the connector sync to another device?
Remote connectors may sync through BoltAI Cloud, but local mcp.json configuration is documented as device-local.
Ready to try it?