GTmetrix
Run real GTmetrix performance tests and get actionable optimization insights directly from your AI assistant.
От сообщества: Добавлен пользователем или импортирован; проверьте владельца перед подключениемРаботаетБез входаГлобальныйБесплатноТолько чтение
Что умеет
- Gtmetrix Get Account Status: Returns the current GTmetrix account status including plan type, remaining API credits, next refill date, and feature access flags.
- Gtmetrix Start Test: Start a new GTmetrix page performance test for a URL. Use this when a user wants to test or improve their website's performance, or reports any of: slow/sluggish loading, layout i
- Gtmetrix Get Test: Get the current status of a started test. The tool long-polls server-side: pass `wait_seconds` (default 25, max 30) and it returns as soon as the test completes or the budget expire
Какие данные видит
Нужен ли аккаунт
Не нужен: сервер работает без входа
Run real GTmetrix performance tests and get actionable optimization insights directly from your AI assistant.
Analyze any webpage's speed and performance without leaving your workflow. Trigger tests, configure browser, location, and device conditions, then get back GTmetrix grades, Core Web Vitals, top Lighthouse issues, and resource breakdowns.
Turn performance data into a natural part of your AI-assisted workflow.
Список инструментов сервера (9)
Технические названия из tools/list. Нужны только разработчикам.
| gtmetrix_get_account_status | Returns the current GTmetrix account status including plan type, remaining API credits, next refill date, and feature access flags. |
| gtmetrix_start_test | Start a new GTmetrix page performance test for a URL. Use this when a user wants to test or improve their website's performance, or reports any of: slow/sluggish loading, layout instability (content jumping), slow server response, poor mobile performance, slow checkout or other flows, high bounce rates, SEO ranking drops, questions about a specific metric (LCP, CLS, TBT, TTFB, etc.), suspecting a recent deploy caused a regression, or wanting to benchmark against competitors. If the user has not provided a URL, ask for it before calling this tool. Concurrency: Basic accounts allow 2 concurrent tests, PRO allows 8. A 429 error means an existing test needs to finish before you retry. Workflow: `start_test` → wait 10 seconds → `gtmetrix_get_test` with `wait_seconds=25` to long-poll server-side; if it returns 'still running', you can re-issue it back-to-back with the same arguments (no need to shell-sleep between calls) until state=complete → `gtmetrix_get_report`. All analysis options are optional; the input schema lists their names, types, and defaults. Most-used options: `location` (test region ID) and `browser` (browser ID) — resolve a region/browser name to its ID via `gtmetrix_get_catalog` (`catalog_id` of `locations` / `browsers`); `report` (default `lighthouse`; also `legacy`, `lighthouse,legacy`, or `none`); `simulate_device` (preset mobile/tablet profile — overrides viewport and `user_agent`; IDs via the `simulated-devices` catalog); `throttle` as `Kbps_down/Kbps_up/latency_ms` (e.g. `1600/768/300` ≈ a slow 3G link; presets via the `throttle-connections` catalog); plus `adblock`, `video` (+0.9 credits), `cookies`, `httpauth_username`/`httpauth_password`, `block_url`/`allow_url`, `browser_width`/`browser_height`/`browser_dppx`, and `retention` (1/6/12/24 months). For the full reference — every option, its default, valid values, and recipes — read the MCP resource `gtmetrix://docs/test-options-guide` (or call `gtmetrix_get_guide` with `guide_id: 'test-options'`). |
| gtmetrix_get_test | Get the current status of a started test. The tool long-polls server-side: pass `wait_seconds` (default 25, max 30) and it returns as soon as the test completes or the budget expires. When it returns 'still running', re-issue the call immediately — do not use shell sleep between calls; the next call will long-poll again. Once the state is complete, use the returned report ID with gtmetrix_get_report to retrieve results. Reports are retained for 1 month by default unless a higher retention was set when the test was started. If the test ends in the `error` state, the response is marked `isError: true` and lists the URI(s) of one or both MCP guides — `gtmetrix://docs/general-test-error-guide` and `gtmetrix://docs/lighthouse-error-guide` — that explain the failure and the recommended next steps. Read the referenced guide(s) before responding to the user (equivalently, call `gtmetrix_get_guide` with `guide_id: 'general-test-error'` or `'lighthouse-error'`). |
| gtmetrix_get_report | Retrieve the report data using the report ID. Reports are the result of a completed test and contain page load information. The result includes a structured hierarchy of web performance metrics: GTmetrix scores, Performance/Structure grades, Core Web Vitals (LCP, TBT, CLS) with qualitative ratings, the top 5 highest-impact Lighthouse issues pre-sorted by engineering priority, a computed Resource Summary, a HAR (waterfall) download URL, and a full Lighthouse JSON download URL. This data may be consumed by human end-users seeking optimization advice, or by autonomous software agents automating development workflows. Tailor the consumption, parsing, and presentation of this data to best serve the immediate objective and technical depth of the consuming entity. Key artifacts in the response: - Resource Summary — per-resource-type request count and transferred / uncompressed byte totals (document, font, image, script, stylesheet, media, other). Pre-computed; no need to re-derive from the HAR. - Top Lighthouse Issues — the 5 highest-impact structural problems Lighthouse found, each with a plain-text description, an Impact label (Critical / Major / Moderate / Minor / Informational), display value, and up to 5 example offending items. Pre-sorted by impact. - HAR (Waterfall) — a HAR 1.2 JSON of every network request from the test, with per-phase timings (DNS, connect, SSL, wait, receive), headers, and sizes. The URL is in the response; download it with curl and inspect locally (it can be many MB, so fetching via MCP is not recommended). If you have no network egress to reach that URL, call `gtmetrix_get_report_har` with the report_slug to fetch it over MCP instead (size-limited). - Lighthouse JSON — the full Lighthouse Report Result with every audit (not just the top 5), full audit details, and metadata. The URL is in the response; download with curl. Resolve unfamiliar audit IDs via the lighthouse-audits catalog. For the full methodology and recommended follow-up flows, read the analysis guide MCP resource `gtmetrix://docs/report-analysis-guide` (or call `gtmetrix_get_guide` with `guide_id: 'report-analysis'`); see also `gtmetrix://docs/har-analysis-guide` (or `gtmetrix_get_guide` with `guide_id: 'har-analysis'`) for HAR field meanings and `gtmetrix://catalog/lighthouse-audits` (or `gtmetrix_get_catalog` with `catalog_id: 'lighthouse-audits'`) to resolve any Lighthouse audit ID. |
| gtmetrix_get_report_har | Fetch the raw HAR (net.har) for a completed report and return it inline. Use this only when you cannot download the HAR URL from `gtmetrix_get_report` directly (e.g. your environment has no network egress to gtmetrix.com); otherwise prefer downloading the URL with your own tools. The HAR is a HAR 1.2 JSON of every network request with per-phase timings, headers, and sizes; Authorization headers and cookies are stripped. By default the HAR is only returned if it is under 256 KB — a multi-MB HAR returned inline can overflow the context window. If it exceeds the limit the tool returns the file size and the download URL instead; you may re-call with `force: true` to return it anyway (use with care). See `gtmetrix://docs/har-analysis-guide` (or `gtmetrix_get_guide` with `guide_id: 'har-analysis'`) for field meanings and analysis recipes. |
| gtmetrix_get_guide | Fetch a GTmetrix documentation guide (markdown). Same content as the equivalent `gtmetrix://docs/<id>-guide` MCP resource; use either surface. You can read a guide before configuring a non-trivial test, interpreting a test error, or analyzing a finished report. Available `guide_id` values: - `har-analysis` — GTmetrix HAR Analysis Guide. Field reference and analysis recipes for the GTmetrix Report HAR (net.har) waterfall. - `test-options` — GTmetrix Test Analysis Options. Reference for every option accepted by the gtmetrix_start_test tool, including defaults, value ranges, related list tools, and credit-cost notes. - `report-analysis` — GTmetrix Report Analysis Guide. Methodology for analyzing a GTmetrix Report: how to present scores and Core Web Vitals, how to read the Top Lighthouse Issues block, and when to download the HAR (per-request timing/headers) vs. - `general-test-error` — GTmetrix General Test Error Guide. How to interpret and respond to general (non-Lighthouse) GTmetrix test errors: server / HTTP status, network connectivity, SSL, DNS, generic timeouts, redirects, and non-HTML responses. - `lighthouse-error` — GTmetrix Lighthouse Error Guide. How to interpret and respond to Lighthouse-specific GTmetrix test errors: missing CPU / network idle, missing FCP / LCP, browser crashed or out-of-memory, debugger-statement pauses. |
| gtmetrix_get_catalog | Fetch a GTmetrix lookup catalog (JSON). Use these to resolve names to IDs accepted by `gtmetrix_start_test` (location, browser, simulate_device, throttle) or to look up Lighthouse audit metadata. Same content as the equivalent `gtmetrix://catalog/<id>` MCP resource; use either surface. Available `catalog_id` values: - `browsers` — GTmetrix Browser Catalog. JSON list of test browsers available to the authenticated user. - `locations` — GTmetrix Location Catalog. JSON list of test locations. - `simulated-devices` — GTmetrix Simulated Device Catalog. JSON list of preset simulated-device profiles. - `throttle-connections` — GTmetrix Throttle / Connection Catalog. JSON list of preset network-throttling profiles for start_test's `throttle` option. - `lighthouse-audits` — GTmetrix Lighthouse Audit Catalog. JSON catalog of Lighthouse audit metadata for the latest supported version. |
| gtmetrix_get_report_history | Fetch the historical performance data for a GTmetrix page. A page is the combination of a URL and its test settings; each time it is tested or monitored a new report is created. Returns all reports for the page in reverse chronological order, including both successful reports (with scores, Core Web Vitals, timing metrics, and a link to each report) and any failed reports. Accepts either a report_slug (to automatically resolve the parent page) or a page_slug directly. Provide exactly one — not both. The response is structured for trend analysis: identifying performance regressions, metric outliers, and patterns over time. Use gtmetrix_get_report with a specific report_id to drill into the full audit details for any anomaly found in the history. |
| gtmetrix_list_pages | List GTmetrix pages for the authenticated account. A page is a URL + test-settings combination; each test run creates a new report. Returns pages with their latest report scores, Core Web Vitals, analysis options summary, monitoring status, tags, and total report count (default 200 pages, max 2000). All filter parameters are optional and combinable: - url: substring match against the page URL - browser: exact browser ID — resolve a name to an ID via gtmetrix_get_catalog (catalog_id='browsers') - location: exact region ID — resolve via gtmetrix_get_catalog (catalog_id='locations') - throttle: exact connection string in 'down_kbps/up_kbps/latency_ms' format — presets via gtmetrix_get_catalog (catalog_id='throttle-connections') - simulate_device: exact device ID — resolve via gtmetrix_get_catalog (catalog_id='simulated-devices') - Boolean flags: adblock, video, stop_onload, anonymize_user_agent, httpauth, cookies, filter, dns — pass true to show only pages with that option enabled - monitoring: filter by monitoring frequency — 'never', 'hourly', 'daily', 'weekly', 'monthly' - private: true = show only private pages (omit to include all pages) - tag: substring match against tag names — shows pages with any matching tag - limit: number of pages to return (default 200, max 2000) For the full option reference read gtmetrix://docs/test-options-guide (or call gtmetrix_get_guide with guide_id='test-options'). Use gtmetrix_get_report_history with a page_slug to drill into trend data for any page found here. |