The agentic platform for website quality.
Webtrax is building an AI agent that keeps your website working. It will check every deploy for broken links, slow pages, security gaps, SEO and accessibility problems, open the fix as a pull request, and re-check once you merge. It starts with the deploy loop on GitHub, Vercel, Netlify and Cloudflare Pages, and with your coding agent over MCP.
One site free forever. Founding members get 40% off Pro or Team for their first year.
How the agent will work: preview, catch, fix, verify. Preview
-
Step 1
Preview
The agent will crawl every preview deploy on Vercel, Netlify or Cloudflare Pages, and production on a schedule.
-
Step 2
Catch
You will get one PR check: "This deploy adds 3 broken links, +400ms LCP on /pricing, and drops a security header." Each finding will be linked to the commit that caused it.
-
Step 3
Fix
The agent will open a pull request with the fix (the corrected link, the redirect, the header config) and the evidence. Nothing changes until you merge.
-
Step 4
Verify
After the merge, it will re-check the affected pages. An issue closes only when the evidence says it is fixed.
One check per deploy. Fixes you can merge.
Every finding will come with its evidence and a proposed fix. "Verified" means the URLs in the fix were checked by our crawler before you saw them.
This is a sample for a made-up store. No scan ran to produce it.
Commit 3f9c2ab, "Redesign footer and hero"
This deploy adds 14 broken links, +1.4s LCP on mobile, and drops 2 security headers.
3 new · 1 fixed · 7 unchanged
-
Fix one footer link to clear 14 broken links
VerifiedHigh impact · effort: 2 minutes
Evidence: 14 pages link to /legal/cookies, which returns 404 since commit 3f9c2ab.
Proposed fix: Redirect /legal/cookies to /privacy#cookies in vercel.json.
View fix PR
-
Three hero images add 1.4s to LCP on mobile
High impact · effort: 15 minutes
Evidence: LCP on / rose from 1.9s (production) to 3.3s (preview); the hero images total 2.1 MB.
Proposed fix: Serve AVIF with srcset; a diff for components/Hero.tsx.
View diff
-
Two security headers missing on every page
VerifiedMedium impact · effort: 1 minute
Evidence: Strict-Transport-Security and X-Content-Type-Options were present in production and are missing on the preview.
Proposed fix: Restore both headers in next.config.js.
View fix PR
What the agent will check
-
Links
Every link, image, script and stylesheet, plus redirects, canonicals and sitemaps. Broken, redirected or timing out, you will hear first.
-
Performance
Lighthouse and Core Web Vitals on the pages that matter, with budgets and regression alerts.
-
Security
A continuous posture grade covering TLS, headers, cookies, mixed content and vulnerable JavaScript.
-
Technical SEO
Canonicals, indexability, noindex surprises, redirects, sitemaps and structured data validity.
-
Accessibility
Automated WCAG checks on rendered pages, with fixes you can ship.
-
Content
Placeholder text, lorem ipsum, stale copyright dates and empty pages, before your visitors spot them.
-
AI-readiness
llms.txt, AI crawler rules and answer-friendly structure, so AI assistants can find and cite you.
-
Monitors
Uptime every minute, certificate and domain expiry, and screenshots when things go wrong.
Use Webtrax from the AI you already use. Preview
Local checks for coding agents
`webtrax mcp` will let your coding agent check localhost for broken links, SEO and accessibility issues before it commits. Open source, and the first thing we ship.
MCP server
Connect Claude, ChatGPT, Cursor or any MCP client, ask "What's broken on my site this week?", and approve the agent's fixes in the chat.
Bring your own key
The agent will run on Anthropic, OpenAI, Google, Mistral, an open-weight model or your own endpoint, with no markup on tokens. Or on included Webtrax AI credits.
API and webhooks
Every action in the dashboard will be in the API, with signed webhooks for every event. n8n at launch; Zapier and Make later.
Planned integrations
GitHub · GitLab · Vercel · Netlify · Cloudflare · WordPress · Shopify · Webflow · Slack · Microsoft Teams · Jira · Linear · Google Search Console · n8n
Vote for yours on the waitlist.
You will stay in charge.
- The agent will change nothing without your approval, unless you switch on auto-fix for low-risk items, with daily limits and one-click undo.
- Your AI key, your provider, your region. We will never train on your data.
- Personal data and secrets will be redacted before anything reaches a model.
- Every agent step will be logged: what it saw, what it proposed, what changed.
Open source at the core.
The Webtrax runner, the engine under the agent, will be Apache-2.0 and is the first thing we ship. You will run it from your laptop, your CI or your coding agent, and connect it to Webtrax Cloud when you want history, alerts, the agent and your team.
Plans
-
Free
$0
One site, watched forever. Ask your AI about it.
-
Pro
$29/mo
Catch it on the preview. Fix it with a PR.
-
Team
$129/mo
The agent fixes it. You approve it.
Questions
What is Webtrax?
An agentic platform for website quality. An AI agent that checks your site on every deploy for broken links, slow pages, security gaps, SEO and accessibility problems, then fixes what it finds as pull requests you review, and proves each fix worked. It is in development: the open source runner ships first.
Is the agent safe to let loose on my site?
It will never change anything without your approval unless you turn on auto-fix for specific low-risk actions. Every change will be logged, undoable and re-verified.
Do I need my own AI key?
No. Every plan will include Webtrax AI credits. Paid plans will also take your own key from Anthropic, OpenAI, Google, Mistral, open-weight providers or your own endpoint, with no markup.
What data goes to the AI model?
Page excerpts and findings, with personal data and secrets redacted. With your own key it will go only to your provider. We will never train on it. Details are in our AI policy.
Can I use it from Claude or ChatGPT?
Yes, through our MCP server, once the cloud beta opens. The free plan will include read-only access. The local MCP server for coding agents comes first, with the open source runner.
When do you launch?
The open source runner and local MCP come first, then a private beta for waitlist members, then public launch. Waitlist members hear about each step first.
What will free include?
One public website scanned weekly at free limits, your top three AI opportunities each month, and read-only MCP and API access. Free forever.
Can it test internal websites?
Not at launch. A runner that lives inside your network, making outbound connections only, comes after launch. Tell us on the waitlist if you need it.
Is the runner really open source?
It will be: Apache-2.0, including the local MCP server and the check rules, published when the runner alpha ships.
Will the prices change?
These are launch prices. Founding members lock in 40% off Pro or Team for their first year.
Be first to try it.
One site free forever. Founding members get 40% off Pro or Team for their first year. One email when we launch, unsubscribe any time.
Join the waitlist