@pulspeed/mcp-server
This MCP server lets AI agents monitor, scan, compare, and diagnose website performance via Pulspeed's 11 tools.
List monitored sites with URLs, scan frequency, and strategy.
Trigger single or bulk PageSpeed scans, either waiting for results or returning async job IDs.
Read Core Web Vitals, score history, and trend metrics over 24h/7d/30d/90d.
Get AI-powered performance recommendations (paid plans).
Compare two snapshots with metric deltas and assessments.
Detect significant performance regressions across scan history.
Check billing-period usage, remaining quota, and plan limits.
Set performance budgets for score, LCP, FCP, TBT, TTFB, and CLS; violations trigger webhooks.
Retrieve raw Lighthouse audit data for specific audit keys.
Get technical SEO audits covering robots.txt, sitemap, meta tags, canonical, structured data, security headers, redirects, hreflang, internal links, and AI crawler access.
Identify sites by
urlorsite_idwhere supported.
Allows retrieving full Lighthouse audit data for detailed performance analysis.
Allows triggering PageSpeed Insights scans to gather performance metrics like Core Web Vitals.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@@pulspeed/mcp-server scan https://example.com"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
@pulspeed/mcp-server
Website performance monitoring for AI agents.
Pulspeed MCP server exposes 11 tools for scanning sites, reading Core Web Vitals, detecting regressions, and getting AI-powered recommendations — all callable by Claude, Cursor, Windsurf, and any MCP-compatible agent.
Quick Setup
Get an API key
Sign up at pulspeed.ai → Settings → API Tokens → create a token.
Claude Code (CLI)
claude mcp add pulspeed -e PULSPEED_API_KEY=your_token_here -- npx -y @pulspeed/mcp-serverClaude Desktop
Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS):
{
"mcpServers": {
"pulspeed": {
"command": "npx",
"args": ["-y", "@pulspeed/mcp-server"],
"env": {
"PULSPEED_API_KEY": "your_token_here"
}
}
}
}Cursor
Edit .cursor/mcp.json (project) or ~/.cursor/mcp.json (global):
{
"mcpServers": {
"pulspeed": {
"command": "npx",
"args": ["-y", "@pulspeed/mcp-server"],
"env": {
"PULSPEED_API_KEY": "your_token_here"
}
}
}
}Windsurf
Edit ~/.codeium/windsurf/mcp_config.json:
{
"mcpServers": {
"pulspeed": {
"command": "npx",
"args": ["-y", "@pulspeed/mcp-server"],
"env": {
"PULSPEED_API_KEY": "your_token_here"
}
}
}
}Related MCP server: SAG RANK MCP Server
Tools
All tools that identify a site accept either url or site_id.
scan_site
Trigger a PageSpeed Insights scan. Creates the site automatically if not yet tracked. Blocks until done (~30–60s) by default.
Param | Type | Required | Description |
| string | ✓ | URL to scan |
|
| — | Default: |
| boolean | — |
|
bulk_scan
Scan multiple URLs at once (max 10). All jobs are dispatched immediately.
Param | Type | Required | Description |
| string[] | ✓ | 1–10 URLs |
|
| — | Default: |
| boolean | — |
|
list_sites
List all monitored sites with latest scores and scan settings.
get_site_metrics
Core Web Vitals for a site with trend analysis and score history.
Param | Type | Required | Description |
| string | ✓ or | Site URL |
| number | ✓ or | Pulspeed site ID |
|
| — | History window (default: |
get_recommendations
AI-powered optimisation recommendations from the latest scan analysis. Requires Developer or Scale plan.
compare_snapshots
Compare two scans with % deltas and a human-readable assessment (🟢 Significant improvement → 🔴 Significant regression).
Param | Type | Description |
| — | Site identifier (compares latest two) |
| number | Specific newer snapshot |
| number | Specific older snapshot |
list_regressions
Find consecutive scans where the performance score dropped significantly.
Param | Type | Description |
| — | Site identifier |
| number | Min score drop to count (default: |
| number | Scan history to analyse (default: |
get_usage
Current billing-period consumption: scans used/remaining, AI analyses, plan limits.
set_performance_budget
Set metric thresholds for a site. Violations trigger threshold.exceeded webhook events on the next scan.
Param | Type | Description |
| — | Site identifier |
|
| e.g. |
|
| Max LCP in ms, e.g. |
|
| Max FCP in ms |
|
| Max TBT in ms |
|
| Max TTFB in ms |
|
| Max CLS, e.g. |
get_raw_audit
Full Lighthouse audit data for a specific audit key — detailed resource-level breakdown.
Param | Type | Required | Description |
| — | ✓ or | Site identifier |
| number | — | Specific snapshot (default: latest) |
| string | ✓ | Lighthouse audit key |
Common audit keys: render-blocking-resources, bootup-time, unused-javascript, uses-optimized-images, network-requests, uses-text-compression.
get_seo_audit
Latest technical SEO audit from CrawlRay: overall score and per-check status with issues. Audits run automatically after performance scans on paid plans.
Param | Type | Required | Description |
| — | ✓ | Site identifier |
Example Prompts
"Scan https://mysite.com and tell me the performance score"
"What's the LCP trend for mysite.com over the last 30 days?"
"My site slowed down today — what changed?"
"Scan all pages: homepage, /about, /blog and compare their scores"
"Set a performance budget: score ≥ 80, LCP ≤ 2500ms"
"Show me the render-blocking resources slowing down mysite.com"Example: Diagnose a regression
User: "My Lighthouse score dropped 12 points overnight. What happened?"
→ list_regressions(url="https://mysite.com", threshold=5)
"🔴 CRITICAL −12 pts: 78 → 66 on 2026-02-22 (snapshot #142)"
→ compare_snapshots(url="https://mysite.com")
"LCP: 2100ms → 3800ms (+81% ✗) TBT: 180ms → 520ms (+189% ✗)"
→ get_raw_audit(url="https://mysite.com", audit_key="bootup-time")
"analytics-v2.js — 1840ms execution. Saves 1.2s if deferred."
Agent: "A new analytics script blocks the main thread for 1.84s,
causing LCP and TBT to spike. Defer or async-load it."Environment Variables
Variable | Required | Default | Description |
| ✓ | — | API token from pulspeed.ai/settings |
| — |
| Override for self-hosted |
Plans
All plans include MCP access. Quotas apply:
Plan | Scans/month | AI Analysis | Webhooks |
Free | 30 | 5/month | — |
Developer ($29) | 500 | Included | 5 endpoints |
Scale ($79) | 2,000 | Included | 20 endpoints |
Development
git clone https://github.com/pulspeed/mcp-server
cd mcp-server
npm install
npm run build # compile TypeScript → dist/
npm run dev # watch mode
npm start # run compiled serverLicense
MIT © Pulspeed
Available Tools
11 toolsbulk_scanA
Scan multiple URLs at once. By default runs scans asynchronously and returns job IDs for polling. Set wait=true to run sequentially and return all results (slower, may take several minutes).
Inputs: urls (required), wait, strategy.
compare_snapshotsA
Compare two performance snapshots for a website. Shows metric deltas with percentage changes and a human-readable assessment. By default compares the latest two snapshots, or specify snapshot IDs.
Inputs: url, site_id, snapshot_id_a, snapshot_id_b.
get_raw_auditA
Get the raw Lighthouse audit data for a specific audit key from the latest snapshot of a site (e.g. 'render-blocking-resources', 'bootup-time', 'unused-javascript'). Returns detailed audit items useful for debugging specific performance issues.
Inputs: audit_key (required), url, site_id, snapshot_id.
get_recommendationsA
Get AI-powered performance recommendations for a monitored website. Requires Pro or Team plan.
Inputs: url, site_id.
get_seo_auditA
Get the latest technical SEO audit for a monitored website (robots.txt, sitemap, meta tags, canonical, structured data, security headers, redirects, hreflang, internal links, AI crawler access) with per-check status and issues.
Inputs: url, site_id.
get_site_metricsA
Get performance metrics for a monitored website, including Core Web Vitals (FCP, LCP, TTI, TBT, TTFB, CLS), trend analysis, and historical data over a specified period.
Inputs: url, period, site_id.
get_usageA
Get current API usage statistics: how many scans and AI analyses have been used this month, remaining quota, and plan limits.
Inputs: no named parameters.
list_regressionsB
Show recent performance regressions for a site: consecutive scans where the performance score dropped significantly. Helps identify when and by how much the site degraded.
Inputs: url, limit, site_id, threshold.
list_sitesA
List all websites being monitored by Pulspeed. Returns site names, URLs, scan frequency, and strategy.
Inputs: no named parameters.
scan_siteA
Trigger a PageSpeed Insights scan for a website URL. Creates the site if not already tracked. By default waits for the scan to complete and returns performance metrics (30-60 seconds). Set wait=false to dispatch asynchronously and get a job_id for polling.
Inputs: url (required), wait, strategy.
set_performance_budgetA
Set performance budget thresholds for a site. When a scan violates these thresholds, webhook events (threshold.exceeded, performance.regression) will fire. Pass null to remove all budgets.
Inputs: cls, fcp, lcp, tbt, +4 more.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 11 tools
Most tools target clearly distinct resources: sites, metrics, regressions, SEO audits, budgets, and usage. Mild overlap exists between scan_site and bulk_scan, and among get_site_metrics, compare_snapshots, list_regressions, and get_raw_audit, but descriptions differentiate them well enough (raw audit per-key vs. aggregate metrics vs. deltas).
Strong verb_noun convention throughout: get_recommendations, list_sites, list_regressions, get_seo_audit, get_site_metrics, set_performance_budget, compare_snapshots. Minor deviations are scan_site (verb_noun but no get/list prefix) and bulk_scan (adjective_noun), which are readable but slightly off-pattern.
11 tools is well within the ideal range and each covers a distinct capability of a performance monitoring service. No tool appears redundant or like filler.
Core read/scan/audit/reporting surface is covered, but there is a notable dead end: both scan_site and bulk_scan return job_id "for polling" yet no tool exists to poll job status or fetch async results. There is also no way to remove/unmonitor a site or update a site's scan frequency/settings, and no tool to read or list webhook/budget configuration beyond setting it.
Maintenance
Related MCP Connectors
Website performance monitoring: scans, Core Web Vitals, RUM data and alerts.
Website checker for AI coding agents. Measures your site in a real browser, with evidence.
Website health, SEO, security-header, SSL and broken-link audits for AI agents.
Live SEO intelligence, rank tracking, site audits, and backlink analysis for AI agents.
Related MCP Servers
- AlicenseAqualityBmaintenanceProvides AI-visibility scoring and site auditing capabilities for websites, enabling agents to check how sites appear in AI engines like ChatGPT and Perplexity, run full SEO/security audits, and monitor changes over time.20715 npm1MIT
- AlicenseNot gradedqualityCmaintenanceEnables AI assistants and coding agents to audit, compare, and optimize websites for performance, SEO, AEO, GEO, and AI crawler accessibility directly from chat or development workflows.MIT
- AlicenseNot gradedqualityCmaintenanceEnables AI assistants to analyze public websites, follow progress, retrieve shareable reports, compare multiple URLs, list checks and recent activity, and surface fixes ranked by impact.Apache 2.0
- AlicenseBqualityBmaintenanceEnables AI coding agents to retrieve real-user performance metrics and 28-day field trends, run Lighthouse lab audits, and access remediation playbooks for LCP, INP, and CLS.7230 PyPIMIT