Introduction to RUMvision MCP
RUMvision supports MCP, which stands for Model Context Protocol. MCP allows AI assistants and other MCP-compatible clients to securely query selected RUMvision data through a controlled set of tools.
Using MCP, AI applications like Claude or ChatGPT can connect to data sources (e.g. local files, databases), tools (e.g. search engines, calculators) and workflows (e.g. specialized prompts)
modelcontextprotocol.io
With the RUMvision MCP, users can ask questions about their organisation’s domains, Core Web Vitals, templates, URLs, browsers, devices, health checks and other real-user monitoring data. This makes it easier to investigate performance issues without manually switching between dashboards and filters.
MCP versus dashboards
The MCP does not replace the RUMvision dashboard. It is an additional way to explore your data conversationally.
Use the MCP when you want to ask targeted questions, compare segments, or quickly understand what stands out in your real-user data. Use the dashboard when you want to monitor trends, inspect charts, validate findings visually, or continue a deeper investigation.
Together, they complement each other: MCP helps you find the direction faster, while the dashboard helps you review and validate the full context.
RUMvision tools
The RUMvision MCP gives AI assistants structured access to real-user Core Web Vitals data, health checks, domain configuration, templates, and issue workflows. Below is an overview of the available tools and what each one can do.
| Tool | Description |
|---|---|
| describe-catalog-tool | Describe the available RUMvision metrics, thresholds, units, dimensions, and filters for a specific domain |
| get-dashboard-link-tool | Generate a direct link to a RUMvision dashboard page for a domain, optionally preconfigured with a metric, device type, percentile, date range, comparison period, grouping, and filters |
| get-diagnosis-guide-tool | Return a metric-specific diagnosis playbook, including recommended dimensions and available breakdown metrics for a domain |
| get-health-checks-tool | List all health checks for a domain, including descriptions, open findings, remediation steps, and help-center links |
| get-issues-tool | Return detailed health-check findings for a domain, with optional filters for status or specific health checks |
| get-top-issues-tool | Return the most impactful open health-check findings for a domain, ranked by priority |
| list-domains-tool | List all domains the connected user can access in RUMvision, including data status, organisation, health-check availability, and MCP access status |
| list-template-rules-tool | List the current URL-matching rules that assign pages to templates for a domain |
| mark-issue-fixed-tool | Mark one or more health-check findings as fixed after a real fix has been applied, so RUMvision can verify them on the next run |
| query-metric-crosstab-tool | Cross-tabulate Core Web Vitals and RUM field data by two dimensions at once, such as template type and device type |
| query-metrics-tool | Query Core Web Vitals and RUM field data for one or more metrics, with optional breakdowns by dimension, filters, percentiles, and comparison periods |
| query-timeline-tool | Track one or more metrics over time, optionally grouped by dimension, to identify when and where performance changed |
| who-am-i-tool | Confirm the authenticated user, accessible domains, and current MCP rate limits |
Start asking questions
Once MCP access is enabled and a connection has been set up, you can start asking questions. For example, a user could ask:
Which domain has the worst LCP this week?
Show me the top Core Web Vitals issues for this organisation.
Which templates have the highest INP on mobile?
MCP access is managed at organisation level. This means you can decide whether nobody, everyone, or only selected users and roles are allowed to query the organisation’s domains through MCP.
Only users with the right permissions can access data through MCP. The MCP does not make domains public and does not bypass the permissions configured in RUMvision.
Usage and token consumption
When using the RUMvision MCP, the external AI client decides how much context it sends to the model and how usage is billed. RUMvision exposes the available MCP tools and data, but token usage and AI-related costs depend on the client, model and subscription you use.
In general, broader questions require more context. For example, asking for a complete organisation-wide analysis may require more tool calls and larger responses than asking for a single metric, domain or template. More context usually means more tokens are processed by the AI client.
How this affects costs depends on the client. Some tools include usage within a subscription or workspace plan, while API-based implementations often charge based on input and output tokens. Always check the pricing and usage limits of the AI client you connect to the RUMvision MCP.
To keep usage efficient, ask specific questions where possible. Mention the domain, metric, period, device type, template or URL you want to investigate. This helps the AI client retrieve only the data needed for the answer.