Developers
Your audit, as your agent sees it
Every audit on your account is one MCP call away. Add it, sign in once, and your coding agent can read every report, compare them, fix what it can, ask us to quote the rest, run the next audit, and even set up a new one with payment in the browser.
What the agent can reach
31 tools: everything the audits know, and everything your dashboard can do.
Across your account5
Every website on the account
list_sitesRead
Each website with its plan, whether an audit is running, how many reports it has and the newest one.
The reports list, across websites
list_reportsRead
Every audit, newest first, with its scores, issue counts and how it compares with the one before.
Issues across audits: open, fixed, regressed
compare_reportsRead
How one website's issues moved over its audits: what is still open, what came back, what was fixed, and what the latest audit did not re-check.
Start the next audit
run_auditReadWrite
Starts one of the plan's extra audits and says how many are left this month.
Ask Refreshly to quote the rest
request_fix_quoteWrite
Sends the issues your agent cannot fix to Refreshly with a note, and confirms we will reply by email.
Inside one report14
The score ring and the eight section grades
get_report_overviewRead
The site, when it was audited, every score, issue counts by severity and the five worst issues by key.
The findings list, worst first
list_issuesRead
One compact row per issue with its key, severity, what was found and the fix in one line. Filter by severity, section, check or page.
A single finding card, opened
get_issueRead
Why it matters, the recommended fix, supporting detail, before-and-after examples, the technical note and every affected URL.
The pages table with speed and issue counts
list_pagesRead
Every page the audit read: status, issue count, worst severity and PageSpeed scores on phone and desktop.
A single page's row, with its issues
get_pageRead
Everything wrong with one URL and its speed measurements, so a fix can be scoped to one template or route.
The AI SEO section: crawler matrix and evidence
get_ai_readinessRead
A verdict per assistant crawler plus the llms.txt, structured data, contact, pricing and freshness evidence behind the score. On an Essentials report the crawler verdicts are the robots.txt half only, and a `withheld` line says what Pro adds.
The accessibility section: every WCAG criterion and its status
get_accessibilityRead
Every WCAG 2.2 A and AA success criterion with its status (fails, potential issue, no automated failures found, or needs a person), the repeated components behind the failures, and the potential issues. Automated testing cannot establish conformance, and the answer says so.
The page speed table and Core Web Vitals
get_performanceRead
The site score, field vitals, every measured page on phone and desktop, what could not be measured, and the performance issues.
The AI visibility section: who the assistants cite
get_answer_visibilityRead
Whether ChatGPT, Google AI Overviews, Perplexity and Claude cite the site for a buyer's questions, question by question with the sources they cited, plus the competitors and platforms cited. It sits outside the letter grade and never judges whether an answer is fair or accurate.
The local search panel: Google Maps, search by search
get_local_searchRead
Where the business sits on Google Maps and in the local map pack for each search, beside each competitor, with the top three businesses Google showed.
The Google Business Profile panel, site and competitors
get_business_profileRead
Each Google Business Profile, the site's first: claimed or not, category, rating and reviews, the newest review, owner replies and what the profile leaves blank. One that could not be found reads as not found, never as missing.
Keyword visibility and the keywords you rank for
get_keywordsRead
Each target keyword's on-page state with its best page and monthly searches, what Google ranks the site and each competitor for, and the searches a competitor ranks for that the site does not.
The competitor analysis: scoreboard and link authority
get_competitor_comparisonRead
The site's score beside each competitor's, section by section, where each side is ahead, how much of the web links to every domain, and any competitor that could not be measured, with the reason.
The strengths list
list_passesRead
Every check the audit found correct, so an agent leaves a working setup alone.
Actions12
Plans
Cancels at the end of the paid period and says when the entitlement ends.
Undoes a pending cancellation so the plan keeps auditing.
Replaces the rivals the next Pro audit compares against and returns what was stored.
Turns the daily site watch on or off for one website.
Reports
Mints a read-only link to one report and says when it expires.
Rotates the report's share secret, so every link handed out stops opening.
Audits and billing
Starts the first Essentials audit of a website never audited before, free and complete, and confirms it is queued.
Creates a Stripe Checkout session for a paid audit; the audit starts when payment completes. For a website that already has a plan, use run_audit; for the first Essentials audit of a new website, use start_free_audit.
A Stripe billing portal link for this account, to open in a browser.
Try each tool
Sample answers from an example account, in the exact shape the server returns.
Across your account5
Inside one report14
Actions12
list_sites
ReadEvery website on the account
The call
▸ list_sites
No arguments beyond the optional report, which defaults to your only website.
What comes back
{ "sites": [ { "siteId": "site_harbourlane", "hostname": "harbourlane.example", "origin": "https://harbourlane.example", "name": "Harbour Lane Coffee Roasters", "plan": { "name": "pro", "status": "active", "nextAuditAt": "2026-10-04T09:00:00.000Z", "extraRuns": { "used": 1, "limit": 3, "resetsAt": "2026-10-01T00:00:00.000Z" } }, "running": false, "reportCount": 5, "latestReport": { "reportId": "rep_09c4", "createdAt": "2026-09-04T09:12:00.000Z", "tier": "pro", "visibility": "full", "score": 71, "grade": "C" } }, { "siteId": "site_northpier", "hostname": "northpier.example", "origin": "https://northpier.example", "plan": { "name": "essentials", "status": "active", "nextAuditAt": "2026-09-22T09:00:00.000Z", "extraRuns": { "used": 0, "limit": 3, "resetsAt": "2026-10-01T00:00:00.000Z" } }, "running": false, "reportCount": 2, "latestReport": { "reportId": "rep_08b6", "createdAt": "2026-08-21T08:40:00.000Z", "tier": "essentials", "visibility": "full", "score": 84, "grade": "B" } } ], "howToUse": "Pass a hostname or report id as `report` to the report tools; call compare_reports with a hostname to see how its issues moved over time." }
Connect it
One line in your terminal, one sign-in in your browser.
Claude Code wants to use your Refreshly account
Signed in as you@example.com
- Read every website and report
- Compare audits and open issues
- Start audits and manage plans on your behalf
- Ask Refreshly for fix quotes
Your browser opens once. Press Allow and every future session is already signed in.
claude mcp add --transport http refreshly https://mcp.refreshly.tech
Then run /mcp in Claude Code and choose Authenticate: sign in to Refreshly once.
Acts as you
Exactly what your dashboard allows: start audits, manage plans, share reports, ask for quotes.
Your account only
The agent reaches the websites and reports on your account and nothing else.
Disconnect anytime
Under AI agent in your dashboard. Access also lapses after 90 days unused.
Then say
One sentence, and the session runs itself.
Use the Refreshly MCP: look at every report on my account from the last 6 months, compare the major issues across them, prioritise, fix what you can in this codebase worst-first, then run a fresh audit and list anything that needs me.
See every website
list_sites
Plans, scores and the newest report per website.
Compare the audits
compare_reportslist_reports
Which issues are open, which came back, which were fixed.
Fix worst first
get_issueget_page
Full detail per issue, then the change in your own code.
Verify and hand back
run_auditrequest_fix_quote
A fresh audit checks the work; a quote request covers what is outside the code.