PageCrawl publishes a plugin with two skills that teach AI agents how to use PageCrawl well. The MCP server gives your assistant PageCrawl's tools. The skills give it the know-how: which tracking mode suits a page, which tool answers a question in one call, when to ask before acting, and how to write a useful summary of what changed.
The plugin and skills are free, open source under the MIT licence, and published at github.com/pagecrawl/skills. They work on every PageCrawl plan, including Free.
What do the skills do?
- pagecrawl helps any agent use PageCrawl from a conversation. It sets up monitors with the right tracking mode, reads what changed with the fewest calls, leads with the changes that matter, triages unreviewed changes with you, checks monitoring health, and sets up webhooks and scheduled reports.
- pagecrawl-api helps when you write code against PageCrawl. It covers creating monitors through the REST API, reading history and diffs, verifying webhook signatures, and pushing values into data sources.
The skills also keep the agent careful. It asks before deleting monitors or clearing changes, never asks for site passwords, treats captured page content as data rather than instructions, and lets PageCrawl do the fetching instead of loading pages itself.
How do I install them?
Claude Code. Inside a session, run:
/plugin install pagecrawl --marketplace pagecrawl/skillsThe plugin includes the MCP connection, so a browser window opens to sign in the first time a PageCrawl tool runs. If you added PageCrawl earlier with claude mcp add, remove that entry with claude mcp remove pagecrawl so the tools are not listed twice.
Claude on the web or desktop, and Cowork. Download pagecrawl-plugin.zip from the latest release, go to Customize > Plugins > Add > Upload plugin, then connect PageCrawl from the plugin's Connectors tab.
Codex. Copy the folders under skills/ into ~/.agents/skills/, then run codex mcp add pagecrawl --url https://mcp.pagecrawl.io/mcp and codex mcp login pagecrawl.
Cursor, GitHub Copilot, Gemini CLI, OpenCode and other agents. Run npx skills add pagecrawl/skills, then add the MCP server in your agent's settings as described in AI Assistants (MCP Server).
Any other agent that reads Agent Skills. Copy the folders under skills/ into its skills directory and connect the MCP server.
What can I ask once it is installed?
- "Watch this product page and tell me when the price changes."
- "Monitor these five competitor pricing pages weekly and tag them competitors."
- "What changed on my monitored pages this week? Lead with what matters."
- "Go through my unreviewed changes with me."
- "Which of my monitors are failing or never change?"
- "Write an endpoint that receives PageCrawl webhooks and verifies the signature."
You do not need to name the skill. The agent loads it when a request is about monitoring pages or about PageCrawl.
What data do the skills send?
None. The skills are instructions: Markdown files with no scripts or background processes. When your agent follows them, it calls PageCrawl's MCP server or REST API with the access you approved, to create monitors and read your monitoring data. Disconnecting the MCP server or revoking your API token in Settings > API removes that access.
Troubleshooting
The agent does not use PageCrawl. Check that the MCP server is connected and turned on for the conversation, then mention PageCrawl in your request.
PageCrawl's tools appear twice. You have both the plugin's connection and an older one. Remove the older entry.
The agent says it cannot find a monitor. Ask it to list your workspaces; the monitor may be in another workspace.