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., "@Website Autonomous MCP Redesign the homepage for mobile and create a Vercel preview."
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.
# Website Autonomous MCP
A production-minded MCP server for Rootellect website design and management workflows.
Project, brand and store resources
Protected path policy
Safe Git task branches and commits
Validation command tools
Playwright browser inspection and screenshots
Vercel preview integration with missing-config safety
JSONL audit logging
Stdio and Streamable HTTP transports
Related MCP server: Coding MCP Server
Redesign the homepage in a premium herbal wellness style. Keep product data, authentication, cart and checkout unchanged. Test mobile and desktop. Create a Vercel preview. Do not deploy to production.
Call prepare_website_edit_task, edit only approved files, then call finalize_website_edit_task.
Version 2 adds specialized website audit agents while preserving all Version 1 tools.
New agents:
design
seo
performance
accessibility
conversion
diagnostics
landing_page
Use run_website_audit for read-only multi-agent inspection. Use prepare_multi_agent_edit_task, validate_multi_agent_edit, and finalize_multi_agent_edit for controlled edits with before/after screenshots, validation, re-audit, optional Vercel preview, and no production deployment.
See:
docs/version-2-agents.md
docs/audit-scoring.md
docs/full-audit-workflow.md
docs/landing-page-agent.md
Analyze caching performance signals.
Inputs: taskId, pathname.
Analyze CTA hierarchy.
Inputs: taskId, pathname.
Analyze design consistency for a page.
Inputs: taskId, pathname.
Analyze font performance signals.
Inputs: taskId, pathname.
Analyze image performance signals.
Inputs: taskId, pathname.
Analyze deterministic page design issues.
Inputs: taskId, pathname.
Analyze product discovery path.
Inputs: taskId, pathname.
Analyze script performance signals.
Inputs: taskId, pathname.
Analyze visible trust signals.
Inputs: taskId, pathname.
Audit conversion flow friction.
Inputs: taskId, pathname.
Audit heading hierarchy.
Inputs: taskId, pathname.
Audit internal links.
Inputs: taskId, pathname.
Audit page metadata.
Inputs: taskId, pathname.
Audit page SEO.
Inputs: taskId, pathname.
Audit structured data.
Inputs: taskId, pathname.
Capture responsive screenshots for an internal page.
Inputs: pathname.
Check visible focus signals.
Inputs: taskId, pathname.
Check mobile touch target sizes.
Inputs: taskId, pathname.
Check Vercel preview deployment configuration.
Inputs: no named parameters.
Compare before and after screenshots for a task.
Inputs: taskId (required), pathname.
Create a safe commit on a task branch.
Inputs: message (required), approvedProtectedChanges.
Create a Vercel preview deployment when configured.
Inputs: no named parameters.
Create an ai task branch.
Inputs: taskName (required).
Return broken image URLs for an internal page.
Inputs: pathname.
Detect horizontal overflow.
Inputs: pathname.
Finalize landing page task validation summary.
Inputs: taskId, pathname.
Validate, audit, compare, commit when safe, and optionally create preview.
Inputs: taskId (required), commitMessage (required), createPreview, approvedProtectedChanges.
Validate, commit, optionally preview, and report a website edit task.
Inputs: taskId (required), commitMessage (required), createPreview.
Generate accessibility fix recommendations.
Inputs: taskId, pathname.
Generate conversion recommendations.
Inputs: taskId, pathname.
Generate a design improvement plan.
Inputs: taskId, pathname.
Generate diagnostic recommendations.
Inputs: taskId, pathname.
Generate performance recommendations.
Inputs: taskId, pathname.
Generate SEO fix recommendations.
Inputs: taskId, pathname.
Generate prioritized recommendations from a saved audit report.
Inputs: taskId (required).
Return changed files.
Inputs: no named parameters.
Return console errors for an internal page.
Inputs: pathname.
Return git diff.
Inputs: no named parameters.
Return git status.
Inputs: no named parameters.
Return homepage source and related brand/store resources.
Inputs: no named parameters.
Return landing page source context.
Inputs: taskId, pathname.
Return network errors for an internal page.
Inputs: pathname.
Get preview deployment status.
Inputs: previewUrl (required).
Return a depth-limited project tree.
Inputs: depth.
Return framework, package manager, scripts and homepage summary.
Inputs: no named parameters.
Return protected path configuration.
Inputs: no named parameters.
Inspect analytics loading signals.
Inputs: taskId, pathname.
Inspect authentication health signals without editing protected code.
Inputs: taskId, pathname.
Inspect browser console errors.
Inputs: taskId, pathname.
Inspect failed network requests.
Inputs: taskId, pathname.
Inspect an internal page for console, network, image and layout issues.
Inputs: pathname.
Inspect payment health signals without editing protected code.
Inputs: taskId, pathname.
Prepare landing page task context.
Inputs: taskId, pathname.
Create branch, capture before screenshots, and lock multi-agent edit scope.
Inputs: instruction (required), agents, pathname, createPreview.
Create and prepare a scoped website edit task.
Inputs: instruction (required), outputMode, targetPage, approvedProtectedChanges.
Push current task branch.
Inputs: no named parameters.
Request gated production deployment.
Inputs: approved (required), previewUrl (required), approvalText (required).
Restore working-tree changes on a task branch.
Inputs: no named parameters.
Return rollback safety status.
Inputs: no named parameters.
Run axe-core and custom accessibility checks.
Inputs: taskId, pathname.
Run deterministic accessibility checks.
Inputs: pathname.
Run lint, typecheck, tests and build.
Inputs: no named parameters.
Run configured homepage Playwright tests.
Inputs: no named parameters.
Run Lighthouse performance audit.
Inputs: taskId, pathname.
Run website lint script when configured.
Inputs: no named parameters.
Run live diagnostics.
Inputs: taskId, pathname.
Run website production build.
Inputs: no named parameters.
Run website typecheck.
Inputs: no named parameters.
Run website unit/test script.
Inputs: no named parameters.
Run deterministic visual checks.
Inputs: pathname.
Run a non-destructive multi-agent website audit.
Inputs: agents, pathname, includeLiveWebsite.
Search safe text files in the website repository.
Inputs: query (required), exclude, include.
Validate page design against brand resources.
Inputs: taskId, pathname.
Validate changed files against permissions and secret scan.
Inputs: approvedProtectedChanges.
Validate landing page prerequisites.
Inputs: taskId, pathname.
Validate changed files and secrets for a multi-agent task.
Inputs: taskId (required), approvedProtectedChanges.
Verify a preview URL returns a healthy response.
Inputs: previewUrl (required).
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 78 tools
Multiple tools have overlapping purposes, such as run_accessibility_check versus run_accessibility_audit, or inspect_page versus get_console_errors and get_network_errors. The lack of clear boundaries between generic audit tools and their specific variants makes it hard to select the right tool.
All tool names follow a consistent verb_noun pattern in snake_case, e.g., get_project_summary, create_commit, run_lint, audit_page_seo. The verbs vary (get, run, audit, analyze, generate) but the structure is uniform.
With 78 tools, the server far exceeds the 50+ threshold for extreme mismatch. Even for a broad 'website autonomous' purpose, the tool surface is excessively granular, with many near-duplicate audit/analysis functions.
The tool set covers the full lifecycle from project analysis, git operations, validation, testing, deployment, and extensive auditing (design, accessibility, SEO, performance, conversion, diagnostics). Minor gaps exist, such as lack of an actual rollback tool (only rollback_deployment returns safety status) and no branch listing/deleting, but these are workable.
Hosted MCP for creating, checking, deploying, and hosting static sites for AI agents.
MCP server for your apps' tools and custom tools, plus hosted AI agents and approval-gated workflows
MCP server for building and testing AI agents with multi-model experimentation and insights.
MCP server that lets your AI assistant build and deploy apps, hosting each one at a live URL.
