Connect the fastry MCP server
MCP is how an AI assistant talks to fastry. After you connect the server, you can ask it to create a monitor and turn PageSpeed on, explain a slow page, hide a finding, follow your Google rankings or send a report. There is nothing to install: the assistant calls fastry over the internet.
Create an API key
The assistant signs in with an API key. API keys are included from the Agency plan. Open API & Webhooks, create a key and copy it. fastry shows the key only once. It starts with fry_.
Give the key the rights the assistant needs. A key that can only read monitors and rankings can look things up. Creating a monitor, a notification or a report needs a key that may write.
Server address and header
The hostname is fastry.com. The full address of the MCP server is https://fastry.com/mcp. It speaks Streamable HTTP, so the assistant sends its requests there directly.
Send the API key in the header Authorization. The value is the word Bearer, a space, and the key. Example: Authorization: Bearer fry_…
Connect Cursor
In Cursor, add a remote MCP server named fastry. Use the address above and the Authorization header. A typical entry looks like this. Replace fry_… with your key.
{
"mcpServers": {
"fastry": {
"url": "https://fastry.com/mcp",
"headers": {
"Authorization": "Bearer fry_…"
}
}
}
}
Connect Claude or ChatGPT
In Claude or ChatGPT, add a remote MCP server. Paste https://fastry.com/mcp as the address and set the header Authorization to Bearer and your API key, the same way as in Cursor.
Some assistants sign you in at fastry instead of a header. You then choose the workspace and the rights. You can remove that access later on the API keys page.
What you can ask for
The cases below name the tools the assistant calls. The fields travel with each tool, so the assistant sees them when it connects. list_monitors is the usual start.
Create a monitor and turn PageSpeed on
Ask for a website monitor for a URL. The tool is create_monitor. psi_enabled turns PageSpeed and performance tracking on, checks the plan limit and starts the first measurement. psi_device is mobile or desktop. scan_after_deploy measures the page again after a deployment. update_monitor changes the same fields later. get_monitor shows whether tracking is on and the last PageSpeed error.
pause_monitor and resume_monitor stop and start the checks. check_now queues one check immediately.
Explain why a page is slow
get_performance returns the Lighthouse scores, Core Web Vitals, CrUX field data, the score history and whether tracking is on. get_lighthouse_audits lists the audits that cost points. get_waterfall and get_resources describe the latest scan: requests, chains and bytes by type. Filmstrip images stay in the app. start_performance_run queues another browser scan and counts toward the daily limit. list_performance_runs lists the scans.
Read findings and hide one
get_insights returns what fastry found on the latest header scan, PageSpeed result and deployment. Active findings are the current ones. Ignored findings stay hidden until they become clearly worse, then they show up as active again. Each finding has a key, a severity (critical, warning or info), a title, and the measured and recommended values.
ignore_insight hides one finding. Pass the monitor and the key, and an optional reason. unignore_insight shows it again. Availability sits beside the findings: get_check_history, get_uptime_history and get_response_times.
Track Google rankings
add_keywords and import_keywords add keywords and start the first measurement. location_code defaults to 2276 (Germany). Other codes are 2040 Austria, 2756 Switzerland, 2840 USA and 2826 United Kingdom. device is desktop or mobile. Mobile needs Agency, Pro, or the ranking package Plus or Profi. Tags are labels you choose.
get_keyword returns the position, the day and week change, the SERP features, the top 10 and up to three competitor domains. get_ranking_history is the daily positions, including SERP features. get_visibility is one monitor's domain: how many keywords you track, the average position, and which keywords moved most. add_competitor and get_competitor_history follow a competitor. list_keywords, update_keyword and delete_keyword list, change or remove a keyword.
Alert, publish and report
create_alert_rule can watch rankings: rank_drop, rank_top10_lost, rank_top10_entered and serp_citation. keyword_domain limits that rule to one domain. list_alert_rules and update_alert_rule manage the rules. list_incidents reads incidents, optionally for one monitor. list_channels and create_channel add email, SMS or a webhook. The address is confirmed as in the app. Slack is connected in the app.
post_status_update writes a public note on an incident. list_status_pages, get_status_page, create_status_page and update_status_page manage the page. create_report builds a client report for a group and a time range. list_reports, get_report and get_report_content read it back as JSON. list_deployments and create_deployment record a release. list_scans, get_scan and start_scan cover a header and HTML scan.
delete_monitor, delete_keyword, delete_alert_rule, delete_status_page and delete_competitor do nothing until confirm is true. The assistant first receives a summary of what would be removed.
Read next
- Create a client report Uptime, performance and rankings for a client, on a schedule or via the API and MCP.
- Create a monitor Website, API, port, ping test and UDP, with regions, rankings and performance.
- Downtime notifications and keyword alerting Email, Slack, webhook and SMS for outages, rankings, SSL and performance.