Author Image

CEO & Co-founder of Visivo

Visivo vs Claude Dashboards: Ask Once or Build to Last

Claude Dashboards, now Claude Data, turns chat questions into dashboards. Visivo is BI as code with a semantic layer. Compare review, refresh, and fit.

Comparison card: Visivo vs Claude Dashboards, ask once or build to last

Claude Dashboards lets you ask a question about your company's data in plain language and get back a live dashboard. Anthropic launched it in beta on October 8, 2026, and on November 10, 2026 it became generally available under a new name, Claude Data. Dashboards made under the old name keep working. This post uses both names, because most people still search for "Claude Dashboards".

Visivo is a different kind of tool. It is BI as code: a YAML project in your repository with a semantic layer, a visual editor, and a cloud that deploys what you commit. So this is less a head-to-head than a question of where each one fits. Anthropic says so too: its launch post says dashboards "work alongside your BI and analytics tools".

Everything said here about Claude Data comes from Anthropic's public pages, linked inline, as of November 10, 2026. Some of its features, such as certified queries, are in beta and may change.

Want to try Visivo while you read? Install the open source CLI with curl -fsSL https://visivo.sh | bash (or pip install visivo), or start free in Visivo Cloud with $30 of credit and no card.

The short version

  • Use Claude Dashboards if you want a fast answer to a one-off question, you already pay for Claude, and nobody needs to own the result next quarter. It needs no project, no YAML, and no install step.
  • Use Visivo if the dashboard is going to last: people will make decisions from it every week, a number must mean the same thing on every chart, and a change should go through review in git before anyone sees it.
  • Use both if your team explores in Claude and then moves the questions that matter into a project with review. That is the pattern this post ends on.

Side by side

VisivoClaude Dashboards
What it is

BI as code: an open source CLI and viewer, plus Visivo Cloud

A Claude feature that builds dashboards and data stories from a chat

Built for

Dashboards a team keeps, reviews, and relies on

"Quick, exploratory questions", in Anthropic's words

How you author

YAML project, a visual editor that saves to that YAML, or an AI agent that writes drafts

Ask in chat; edit titles, charts, or a chart's SQL query directly

Shared definitions

Semantic layer: metrics, dimensions, and relations (declared joins)

Certified queries (beta), certified in Claude or reused from dbt™, Looker, or Cube over MCP

Where the dashboard lives

YAML files in your git repository, deployed to Visivo Cloud or built as a static site

An artifact in Claude, with version history and restore

Review before publish

Drafts, then commit; each Visivo Cloud branch is tied to a git branch

Version history with restore; certified queries can go through GitHub pull requests

Data refresh

You run visivo run from your own CI job or cron

Re-queries when someone opens it and its data is due for a refresh; no fixed schedule

Sources

Ten source types, including Snowflake, BigQuery, Postgres, Redshift, ClickHouse, DuckDB, and CSV or Excel files

Warehouses such as Redshift, BigQuery, ClickHouse, Databricks, and Snowflake, plus connected apps such as Salesforce

Testing and alerts

Data tests in the project, alerts to Slack or email

A Test button checks a query before it is certified; no data tests or alerts in Anthropic's public docs

Price

Free CLI; free cloud tier; Platform $500 a month, flat

Included in paid Claude plans; counts toward plan usage limits

What Claude Dashboards does well

Give it credit, because it is good at the thing it was built for. You connect a warehouse or an app, ask "how do this week's signups compare with last month's?", and get a dashboard in the same conversation. According to Anthropic's help article:

  • Each chart shows its query. Claude writes a SQL query for each chart, and you can open it to check what is counted. Each chart also shows when its data was last refreshed.
  • You can edit it directly. Ask Claude for a change, or edit a title, a caption, a chart, or a chart's query yourself.
  • It keeps versions. You can preview an earlier version and restore it.
  • It does more than dashboards. Claude Data also makes scrolling data stories, short narrated videos, and written analyses from the same data.
  • There is no install step. If your company already pays for Claude, the feature is part of the plan.

For a question you will ask once, that is hard to beat. Nobody should write a YAML project to settle a hallway argument about last week's signups.

Definitions: certified queries vs a semantic layer

Both tools can draw a chart. The hard part is making sure that "revenue" means the same thing on every chart, for every person, every time. Each tool answers that in a different way.

Claude Data uses certified queries, a beta feature. A data team approves a query, Claude can reuse it when it builds new dashboards, and charts built on it carry a badge. The team can certify a query in Claude, or bring definitions it already has in dbt™, Looker, or Cube through existing MCP connectors. Editing a certified query removes its certification. Anthropic's pages describe certified queries, not a layer of metrics, dimensions, and joins. In Claude Data, the unit a team approves is the query.

Visivo has a semantic layer in the project itself. Metrics are aggregate definitions such as revenue. Dimensions are row-level fields such as order month. Relations declare how models join. Because these parts compose, one insight (the bar chart defined in the YAML below) can put revenue from orders against a segment from customers, and Visivo writes the join from the relation. Nobody has to approve a new query for each new combination:

name: revenue-project

models:
  - name: orders
    sql: SELECT * FROM orders
    metrics:
      - name: revenue
        expression: SUM(amount)
    dimensions:
      - name: order_month
        expression: DATE_TRUNC('month', ordered_at)
  - name: customers
    sql: SELECT * FROM customers

relations:
  - name: orders_to_customers
    join_type: left
    condition: ${ref(orders).customer_id} = ${ref(customers).id}

insights:
  - name: revenue-by-segment
    props:
      type: bar
      x: ?{ ${ref(customers).segment} }
      y: ?{ ${ref(orders).revenue} }

Certified queries are a sensible design for a chat tool, and the dbt™, Looker, and Cube import means a team with an existing semantic layer can bring it along. If you do not have one yet, Visivo gives you one, and the Metric reference covers the syntax.

Where the dashboard lives

A Claude dashboard is an artifact in Claude. It has version history inside Claude, and you share it from Claude. Anthropic's pages do not describe exporting a dashboard as code into your repository.

A Visivo dashboard is a set of files in your git repository. It goes through the same pull request review as the rest of your analytics code. You can deploy it to Visivo Cloud with visivo deploy, or build a static site with visivo dist and host it anywhere. If you stop using Visivo Cloud, the project is still yours, in plain YAML.

That difference matters most for dashboards that outlive the conversation that made them. A board metric, a weekly ops review, or a customer-facing number should have an owner, a history of who changed what and why, and a review step before a change goes live.

When the question gets bigger

Anthropic places Claude Data next to your BI tool, not in place of it. Its help article says: "When a question needs deeper analysis, send the dashboard to another tool and continue there." At launch, the handoff destinations were Amplitude, Grafana, Hex, Mixpanel, Omni, Perplexity, PostHog, and Sigma. The November announcement lists Amplitude, Mixpanel, Omni, PostHog, and Sigma as its data partners.

Those are hosted apps, mostly analytics and BI tools. Visivo is not on the list, so you cannot send a Claude dashboard to Visivo with one click today. What you can do is rebuild the questions that matter as a Visivo project, which the next section covers.

Agents that write drafts you review

Visivo's position is that AI should write dashboards, and that those dashboards should land somewhere with rules.

Visivo's built-in agent works inside the project. It reads existing objects, validates a proposed change against the project schema, and writes it as a draft. A person sees the pending changes and commits or discards them. The agent cannot commit, deploy, or delete on its own.

Local visivo serve also exposes an MCP server at /api/mcp/. Claude Code, or another MCP client, can use it to read your models and metrics and write new insights, charts, and dashboards as drafts. The agent works on top of your semantic layer, so the numbers it charts use the same definitions as every other chart in the project.

So a practical split looks like this:

  1. Explore in Claude Dashboards. Ask quick questions, throw most of the answers away.
  2. Keep what lasts. When a question turns into a weekly habit, define its metrics and dimensions in Visivo, with Claude Code doing the typing over MCP if you like.
  3. Review and ship. The change goes through a pull request and deploys like any other code.

Who should pick Claude Dashboards

Claude Dashboards is the better choice if:

  • The question is one-off. You want an answer now and do not need the dashboard next month.
  • You already pay for Claude and do not want another tool. The feature is part of your plan.
  • Your data is in an app as well as a warehouse. Claude Data reads apps connected to Claude, such as Salesforce. Visivo connects to databases and files, not SaaS apps.
  • You use Databricks. Claude Data lists it as a warehouse source. Visivo does not have a Databricks source today.
  • You want a story or a video, not a dashboard. Scrolling data stories and narrated shorts are things Visivo does not make.

Who should pick Visivo

Visivo is the better choice if:

  • The dashboard is going to last and needs an owner, a review step, and a history in git.
  • You care most about consistent numbers and want metrics, dimensions, and declared relations in one layer, so new combinations do not need a new approved query.
  • You want agents to write dashboards without letting them publish on their own. Visivo's agent and MCP tools write drafts for a person to commit.
  • You want to test the data behind a dashboard and get an alert in Slack or email when a test fails.
  • You want your dbt™ models, not only certified queries. Visivo reads dbt™ models as a source and builds metrics on top of them. Claude Data can reuse dbt™ definitions over MCP; Visivo works from the models themselves. See sources and dbt™.

Try Visivo

Install the CLI and build your first dashboard locally:

curl -fsSL https://visivo.sh | bash
visivo init
visivo serve

Or start free in Visivo Cloud: no card, $30 of credit to begin, and Platform at $500 a month when your team is ready. The pricing page has every tier, and the product overview walks through each capability.

Install command copied