Update Existing Flourish Embeds Without Replacing Them

The Flourish Connector can apply new data to several existing charts from one conversation in Claude, ChatGPT or another MCP-compatible tool. Supply the existing visualization IDs and clean source data, inspect every result, then republish each visualization from the Flourish editor. Keeping the same IDs means the existing embeds do not need to be replaced.

The editor is the publication gate. The Connector can create or edit a visualization, but publishing still happens in the Flourish editor, so Connector edits can be reviewed before they go live.

Mark the three workflow gates to see whether you are ready to update or republish.

Prepare an Update Manifest

Make a small control table separate from the chart data. It should map every source file to the existing visualization it is intended to update.

Chart key	Visualization ID	Expected rows	Key / period
brand_health	12345678	13	month / 2026-08
channel_reach	23456789	8	channel / 2026-08
share_of_voice	34567890	6	brand / 2026-08

These values are fictional. Use the ID for each existing Flourish visualization and a readable chart_key that helps prevent a dataset from being attached to the wrong chart.

Prepare one clearly named CSV or spreadsheet tab per chart. Preserve the column names and data types expected by the existing project unless the chart itself is meant to change. Remove notes, subtotals, merged headings and unused columns.

Record the expected detail-row count, key field and reporting period in the manifest. These values give the reviewer concrete checks after the Connector has loaded the data. A row count alone is insufficient because the correct number of rows can still contain duplicated categories or the wrong reporting period.

Only supply data approved for the intended publication and for processing by the connected AI service. Flourish says visualizations created through the Connector retain normal Flourish access controls, but the chosen AI tool processes the data supplied to it.

If percentages are involved, settle whether the source stores 42% as 42 or 0.42 before updating. The percentage export checklist covers this common source of chart errors.

Preserve a Recovery Copy Before Editing

Before authorizing changes, preserve the prior source files and make a recovery copy of any critical project. Flourish’s current help documentation says Connector edits to an existing project have no version history or undo (Flourish Connector help).

Keep the recovery copy separate from the project being updated. Also retain the exact approved files sent through the connected AI tool. If validation finds a problem, those files establish whether the error came from the source, an automated transformation or the chart configuration.

The Connector respects Flourish permissions: it can change a visualization when the authenticated user has edit access. Confirm that the connected Flourish account has access to every ID in the manifest before starting a multi-chart update.

Ask the Connector to Update Existing IDs

Flourish demonstrates a workflow that supplies three existing chart IDs and new monthly data in one conversation (six Flourish AI workflows). A controlled prompt can look like this:

Update these existing Flourish visualizations; do not create replacements.

• brand_health, visualization ID 12345678, use brand_health_2026-08.csv
• channel_reach, visualization ID 23456789, use channel_reach_2026-08.csv
• share_of_voice, visualization ID 34567890, use share_of_voice_2026-08.csv

Preserve each chart’s template, colors, title structure, annotations and data bindings. Replace only the chart data and reporting-period text. For each visualization, state the row count you used and return its edit link. Do not publish.

Specifying existing IDs distinguishes an update from the creation of separate visualizations. A newly created visualization would have a different ID and would not be the visualization referenced by the existing embed.

The Connector uses Model Context Protocol, or MCP, to let a compatible AI tool work with Flourish. It can transform data, configure a chart and return an edit URL rather than a static image.

Treat any automated transformation as unverified. Flourish warns that automated reshaping can produce incorrect values, especially with large or complex datasets, and that an incorrectly inferred column type may need manual correction in the editor (Connector help).

If the returned row count differs from the manifest, stop before republishing. Determine whether the Connector filtered rows, treated a heading as data, aggregated records or loaded the wrong file. Do not revise the expected count merely to match the result unless the approved source itself has changed.

Validate Every Updated Chart

Open each returned edit link and compare the visualization with its approved source file. Review the charts individually even when the Connector updated them in one conversation.

1. Identity: Confirm that the visualization ID matches the manifest.
2. Rows: Confirm that the loaded detail rows match the expected count.
3. Keys and periods: Check that categories, dates and reporting periods are complete and unique where required.
4. Totals and extrema: Independently recalculate totals, minimums and maximums from the approved source.
5. Types and formats: Verify that dates remain dates, numbers sort numerically, and percentages and currencies use the intended scale.
6. Bindings: Confirm that labels, values, tooltips and filters still use the correct columns.
7. Editorial elements: Update the title, subtitle, source, notes, annotations and accessibility description for the new period.
8. Responsive display: Inspect narrow and wide layouts, including hover or tap interactions.

A correct-looking chart can still have the wrong bindings. Check the editor’s data and binding settings rather than relying only on the rendered marks. Pay particular attention when old and new files contain similar column names or when an automated transformation changed a type.

Use the Flourish Assistant for editor-level refinements such as sorting, colors or a source line. Unlike Connector changes to a project, Assistant changes map to visible editor settings and are reversible, according to Flourish’s launch announcement.

Record the result of each validation against its manifest row. A failed chart should remain unpublished while the other visualizations are reviewed; a successful batch request does not establish that every result is correct.

Republish the Existing Visualization

Once a visualization passes review, republish that existing visualization from the Flourish editor. A standard Flourish embed identifies its visualization through a value such as data-src="visualisation/4853699" (Flourish embedding documentation). If the visualization ID is unchanged, the page’s embed reference does not need replacement. Republishing updates the content served for that visualization.

Load the actual article or dashboard rather than stopping at the editor preview. Confirm the new reporting period, values, tooltip content and mobile layout. If an old version appears, reload without cached content before treating it as a failed update.

Update the surrounding page metadata as part of the same release. A chart showing August values beside a note that says “Updated July” leaves readers unsure which date is authoritative. Place source, coverage and update dates beside the visualization using a consistent public-data metadata block.

Keep the manifest, approved source files and validation record together. The Connector reduces repetitive work across several visualizations; the manifest, recovery copy and manual publish gate keep each update traceable and reviewable.