Can you provide an example of a Claude dashboard?

Why Everyone's Asking for Claude Dashboard Examples

Forty percent of US employees now use AI at work, a figure that has doubled since 2023. This rapid adoption creates a specific operational challenge: teams want live visibility into their data but lack a concrete example of what a "Claude dashboard" actually looks like in practice.

The confusion stems from two distinct dashboard modes that operate under different technical architectures. Understanding the difference is essential for anyone trying to build a useful interface without writing code.

First, there are Claude Dashboards, a beta feature available on Pro, Max, Team, and Enterprise plans. In this mode, users connect a data warehouse and ask questions in plain language. Claude then writes a SQL query for each chart and runs it against the source. As Anthropic’s official documentation explains, these are designed for quick, exploratory questions like comparing this week’s signups with last month’s.

Second, there are Claude Code artifacts, which function as live, shareable HTML pages. These are not static images but dynamic templates that pull data in real-time. When you share an artifact, it does not hold credentials. Instead, each viewer authenticates their own MCP connectors, ensuring they only see data they have access to.

A practical example of this second mode involves building a status page. You might prompt Claude Code: "Build a dashboard artifact showing open GitHub issues for this repo, recent Sentry errors from the last 24h, and today's deploy status." As described by AlphaSignal, Claude writes the HTML, wires in the data bindings, and publishes it to a private URL.

This approach offers significant speed advantages, but it has clear limitations. The artifact is a single static HTML page with a 16 MiB size limit and no backend support. It is not a replacement for enterprise BI tools like Tableau or Looker, which offer robust audit logs and role-based access controls.

For teams looking to extend these capabilities, the underlying mechanism is the Model Context Protocol (MCP). Anthropic’s directory currently lists 554 integrations across 30 categories. However, selecting the right tools requires careful vetting. SkillzBench is building a free directory and MCP connector to help users verify which agent skills are secure and effective before deployment.

Key distinctions between the two dashboard modes include:

• Data Refresh: Claude Dashboards update based on SQL queries, while artifacts pull live data at view time.
• Security Model: Artifacts isolate credentials per viewer, whereas standard dashboards rely on the creator’s connection.
• Sharing Capability: Artifacts can be shared via private URLs where viewers authenticate independently; standard dashboards start private and are shared via a button.

Knowing which mode fits your workflow prevents wasted effort. If you need quick answers from a warehouse, use the standard dashboard. If you need a live status page for your team, build a Claude Code artifact. The choice depends on whether you want a snapshot of historical data or a live view of current operations.

Two Real Claude Dashboard Examples, Start to Finish

Two very different dashboard examples show what "asking Claude for a dashboard" actually looks like in practice — one turns plain English into SQL charts, the other builds a live, shareable web page wired to your tools.

Example 1: Plain-language questions into SQL charts. Anthropic's official documentation describes Claude Dashboards as a beta feature on Pro, Max, Team, and Enterprise plans that turn questions about company data into dashboards. You connect a data warehouse, ask in plain language, and Claude writes the queries and builds the dashboard. Anthropic's own example prompts include "Build a dashboard of this quarter's revenue by region," "Show weekly signups for the last six months, split by plan," and "Make a dashboard of my open Salesforce opportunities by stage and owner," with the guidance to "say what you want to know, not just what to chart."

Under the hood, each dashboard is an artifact built from the Dashboards template, and Claude writes a SQL query for each chart, running it against the connected source. Supported sources include Snowflake, BigQuery, Amazon Redshift, ClickHouse, Databricks, and apps like Salesforce. Every chart displays its query and last refresh time, dashboards save to the Artifacts tab, start private, and can be shared with a Share button.

Example 2: A live dashboard wired to GitHub, Sentry, and deploys. The second example comes from Claude Code. A widely shared prompt pattern asks: "Build a dashboard artifact showing open GitHub issues for this repo, recent Sentry errors from the last 24h, and today's deploy status. Keep it updated as you investigate. Use the GitHub and Sentry MCP connectors." Claude writes the HTML page, wires in MCP data bindings, and publishes it to a private URL on claude.ai.

What the user sees is a single static HTML page — capped at 16 MiB, with no backend — that calls MCP connectors at view time. Each viewer authenticates their own connectors on first interaction, so the artifact never holds credentials and viewers only see data they have access to. Anthropic's connector directory lists 554 MCP integrations across 30 categories, and each publish creates a new version in history.

Both examples share one lesson: the dashboard is only as good as the connections behind it. Claude Dashboards are built for quick, exploratory questions — how this week's signups compare with last month's — and Anthropic recommends sending deeper analysis to another tool. AlphaSignal's assessment is blunt: these are not replacements for Tableau, Looker, or Power BI, which offer role-based access controls, audit logs, and enterprise governance Claude doesn't yet have.

That connection layer is also where agent skills come in. With published research finding that roughly 26% of agent skills carry at least one vulnerability, choosing connectors and skills carefully matters — which is why SkillzBench is building a free directory of verified, benchmarked skills plus a free MCP connector, so you decide exactly what your agents can see. Everything is free, and the waitlist is open now.

• Claude Dashboards: plain-language question in, SQL-driven chart out
• Claude Code artifacts: live HTML dashboards wired to MCP connectors
• Per-viewer authentication keeps credentials out of the artifact
• Both suit exploratory questions, not enterprise BI governance

How MCP Connectors Power Every Live Dashboard

The magic of a live Claude dashboard isn't the chart — it's what happens the moment someone opens it. Claude Code artifacts can call MCP connectors at view time, pulling fresh data for each viewer using their own credentials, so ten people can share one dashboard URL and each see only the data they're authorized to see.

The scale of what can be wired in is striking. Anthropic's connector directory lists 554 MCP integrations across 30 categories, including tools like GitHub, Slack, Jira, Sentry, Google Calendar, and Asana, according to AlphaSignal's coverage. That breadth is why a single prompt like "show open GitHub issues, recent Sentry errors, and today's deploy status" can produce a working dashboard rather than a static mockup.

The security model is equally deliberate, and worth understanding before you share anything:

• The artifact never holds credentials — no tokens or keys are baked into the page itself.
• Each viewer authenticates their own MCP servers on first interaction, so access follows the person, not the link.
• Viewers only see data they already have permission to access in the underlying tools.
• Org admins can enable or disable artifact MCP access at the organization level.

AlphaSignal describes the result as "a template each viewer hydrates with their own connected data" — the dashboard is the shell, and your authenticated connections fill it in. Even the cost follows the same pattern: usage counts against each viewer's own Claude subscription limits, not the creator's.

There are honest limits to keep in mind. Artifacts are single static HTML pages capped at 16 MiB with no backend, and CSP rules block external requests at render time. Anthropic's own documentation positions dashboards for quick, exploratory questions — not as a replacement for enterprise BI tools like Tableau, Looker, or Power BI, which offer governance features Claude doesn't yet have.

For teams wanting to go further, the extension point is the same one that powers these dashboards: MCP. SkillzBench is building a free MCP connector alongside a directory of agent skills, letting you choose exactly which skills and tools your agents can wire into workflows like these — with every skill pinned to a specific version, security-scanned, and benchmarked before it reaches your agent. The waitlist is open now for anyone who wants to be first in line when it launches.

Limits, Plans, and Honest Expectations

Before you build anything on Claude Dashboards, know exactly what they are — and what they are not. Anthropic's own documentation describes them as tools for quick, exploratory questions, like how this week's signups compare with last month's. That framing matters more than any feature list.

Start with availability, because it gates everything. Dashboards are a beta feature on Pro, Max, Team, and Enterprise plans, and on Enterprise they stay off until an owner enables them under Organization settings > Artifacts. Claude Code artifacts are stricter still — they require Team or Enterprise plans. If you are on a solo Pro subscription, some of what follows simply is not open to you yet.

Then come the hard technical ceilings, which AlphaSignal's breakdown lays out plainly:

• A single static HTML page, capped at 16 MiB rendered size
• No backend of any kind
• A content security policy that blocks fetch, XHR, and WebSocket calls at render time
• No public sharing when live MCP data is involved
• Unavailable under CMEK, HIPAA, or ZDR compliance configurations

That combination is deliberate, not accidental. Because the artifact never holds credentials and each viewer authenticates their own MCP connectors, the page stays a template rather than an application — what Anthropic calls "a capture of work, not an application."

So set expectations accordingly. AlphaSignal states directly that Claude artifacts are not a replacement for Tableau, Looker, or Power BI, which offer role-based access controls, embedded analytics, audit logs, and enterprise governance Claude does not yet have. What you get instead is zero-infrastructure speed — a dashboard in minutes, with no pipeline to maintain.

For teams extending agents beyond dashboards, the same MCP plumbing is where skills come in. SkillzBench is building a free, independent directory of agent skills plus a free MCP connector, so you can choose exactly which skills and tools each agent sees — with security scanning and benchmarking in development. The waitlist is open.

Extend Your Dashboard Workflow with Agent Skills

Your dashboard is live and answering questions — but dashboards answer; skills act. The natural next step is letting agent skills handle the investigation and follow-up work behind the charts, so your dashboard becomes the starting point of a workflow rather than a static snapshot.

This matters more as adoption accelerates. Roughly 40% of US employees now report using AI at work, up from 20% in 2023, according to Anthropic's economic index — and more of that work is shifting from answering questions to completing tasks. Agent skills (SKILL.md instruction packages that work with Claude Code, Codex, Cursor, Gemini CLI, and chat-app agents) are how you delegate those tasks.

But there's a catch worth taking seriously. Published research cited by SkillzBench found that roughly 26% of agent skills carry at least one vulnerability. Installing a random skill from GitHub into an agent that touches your GitHub, Sentry, or Salesforce connectors is a real security decision, not a convenience toggle.

That's the gap SkillzBench is being built to close. It's a free, independent directory of agent skills — currently in pre-launch, with a live waitlist — that continuously crawls and dedupes skills from GitHub, then puts each one through a verification pipeline before you ever see it:

• Every skill version is pinned to a commit SHA, license-checked, and security-scanned
• Each skill is benchmarked in a sandbox with the same model, with and without the skill, and the resulting Lift score is published in percentage points — shown separately from ratings, with no money ever touching the scores
• A trust score from 0–100 combines security, quality, license, maintenance, and community signals
• Upstream changes are tracked, re-scanned, and re-benchmarked, with alerts when a skill you rely on shifts

The workflow fits neatly with how these dashboards already operate. A Claude Code artifact pulls live data through MCP connectors — Anthropic's directory lists 554 integrations across 30 categories — and each viewer authenticates with their own credentials, so the artifact never holds them. SkillzBench extends that same principle to skills: its free MCP connector plus per-agent loadouts mean you toggle exactly which skills and tools each agent can see, and nothing more.

Think of it in terms of the dashboard example itself. A dashboard showing open GitHub issues, recent Sentry errors, and deploy status tells you something needs attention. A verified skill package can be the thing that triages the issue, checks the error pattern, and drafts the follow-up — while the dashboard keeps watch.

Everything on SkillzBench — the directory, the MCP connector, benchmarks, and alerts — will be free, supported by advertising around its blog rather than user fees. If you want your dashboards to do more than display data, join the waitlist and get verified skills the moment they're proven.