Flourish MCP server icon

Flourish

by Flourish

HIPAA CompliantSOC2 ReadyISO 27001 Ready
Productivity10 tools

Create and edit Flourish data visualisations from an AI agent. Anthropic lists 10 tools; Flourish publishes no tool list but states plainly that publishing stays in the editor. Two OAuth scopes, and no tool reaches a public URL.

Verified connector

Listed by Anthropic as a partner connector in its Connectors Directory.

Connection checked by Agentman on .

Anthropic states this reflects the level of review a connector received, not a security audit.

Connect Flourish via MCP

https://app.flourish.studio/mcp

Works in any MCP-compatible client. In Agentman Studio it is one click — no config file to edit.

Flourish Tools & Capabilities (10)

Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.

Limits

  • The connector cannot publish, and that is the design. Nothing here reaches a public URL, an embed code or a downloadable HTML bundle. If you want the chart live, you open the editor. This is a limit and a safeguard at once.
  • No version history and no undo. Flourish states plainly that MCP edits to an existing project have neither, and calls it a beta-phase item it is monitoring. flourish_duplicate_visualisation is the only recovery mechanism, and only if you use it first.
  • Flourish publishes no tool list for this server. The 10 names come from Anthropic's directory. We could not diff them against a vendor enumeration of names because none exists — but Flourish's published statement about what the connector *cannot* do is a stronger constraint than a name list, and the advertised OAuth scopes independently agree with it.
  • We could not read tool schemas or safety annotations. The endpoint returned 401 to an anonymous request, so no parameter-level detail is published here.
  • Data reshaping can silently corrupt values. Flourish warns that when the agent reformats data — long to wide, for example — it "can occasionally produce incorrect values, especially with large or complex datasets", and asks you to double-check. This is the failure mode to watch, because a wrong chart looks exactly like a right one.
  • The agent cannot fix a column type. Flourish states that if the agent infers a column's type incorrectly, treating dates as text for instance, it cannot correct that programmatically. You must change the data type in the Flourish editor.
  • The agent cannot change your theme. Flourish states a default theme is applied to newly created visualisations as in the editor, but the MCP "can't change or switch themes".
  • Direct uploads cap at 1,000 rows or 1MB. Larger datasets need another route; Flourish says the agent will guide you to alternatives.
  • AI features are off by default on company accounts and need both an admin action and a per-user one before the connector works at all.
  • Connections are per user. No shared team connection exists.
  • A published chart can carry its data to strangers — but only once a person publishes it. Flourish's publishing flow offers data download links, and on Free accounts published projects are duplicable by default, copying the project and its data. Relevant when *you* publish what the agent built, not when the agent builds it.
  • The directory's documentation URL redirects. Anthropic lists .../16487012057999-Flourish-MCP-connector-an-intro; the article's canonical slug is .../16487012057999-Flourish-Connector-an-intro. Same article, renamed. The help centre's Cloudflare layer returns 403 to plain clients, so we read it through Zendesk's public help_center article API on the same host.

Frequently asked questions

No. Flourish states on its Connector product page that publishing still happens in the Flourish editor so you can review the visualisation before it goes live. No tool among the ten carries a publish, share or embed verb, and the server advertises only visualisation read and write scopes. The agent returns an edit link, and a person clicks Export and publish.

Exactly two, and they are real application scopes. The server's RFC 9728 descriptor advertised visualisation:read and visualisation:write on 2026-08-23, and its authorization server metadata advertises the same pair. The boundary falls between reading a chart and changing one, which is the useful place. It does not separate creating a new chart from overwriting an existing one.

Yes, in both directions. flourish_get_visualisation_data reads the dataset out of an existing visualisation into the agent's context, and flourish_update_visualisation_data writes a dataset in. Flourish confirms your AI tool processes the data you provide. Nothing becomes public, because a chart stays private until a person publishes it in the editor.

It cannot delete one, and it can overwrite one. No tool carries a delete, remove or archive verb, and Flourish's project-deletion flow has no counterpart in the listing. But three update tools rewrite an existing visualisation's data, data bindings and settings, and Flourish states there is no version history or undo for connector edits.

Because AI features are off by default on Flourish company accounts. Flourish states this is intentional and does not depend on which paid plan you are on. Your company admin must click Enable Connector in company settings, and your individual account needs the feature enabled too. Personal accounts enable AI-powered features in account settings.

No. Flourish states that for the Connector, token cost sits entirely with your chosen AI tool rather than with Flourish, and that the Connector does not consume Flourish Assistant credits. The AI processing happens in Claude or ChatGPT, so usage is billed through that tool. The Connector is available to all Flourish users.

Up to 1,000 rows or 1MB when you upload data directly, per Flourish's help documentation, with the agent guiding you to alternatives above that. Because MCP is an open protocol, your AI tool can also pull data from other connected sources such as Google Sheets, databases or APIs and pass it straight into Flourish.

No. Flourish states the MCP is authenticated on a single user level, so each team member connects their own account. The connector then respects your existing Flourish permissions: if you have edit access to a visualisation the agent can change it, and if you do not, it cannot. Charts inherit the editor's access controls.

Sources

Use in Agentman

Connect once and your agents call these tools on their own — on a schedule, in a workflow, with nobody at the keyboard.

Open in Agentman Studio

Server Info

Category
Productivity
Developer
Flourish
Tools
10
Domain
app.flourish.studio

Using Claude Desktop or another MCP client? Setup docs — the connection URL above works anywhere.