How I Automated Finances with Claude AI

Install a Claude Cowork plugin and build an auditable monthly finance dashboard.

Introduction

30 Second Summary

A bank export can show every purchase while still leaving the monthly decisions buried in rows. Repeating the review by hand makes each month's result harder to trust.

In this project, you will install or build the Money Dashboard plugin in Claude Cowork to turn synthetic transaction CSV files into a repeatable monthly review. The finished workflow keeps each figure traceable to its source while guiding your monthly decisions.

What You'll Build

Picture opening your Money Dashboard to see February beside January with the next subscription decision already surfaced.

By the end of this project, you'll have:

  • An interactive monthly dashboard where you can switch between January and February to inspect cash flow, category totals, and transaction rows.
  • A repeatable import workflow that remembers approved merchant categories. It excludes an exact duplicate before that row can inflate spending.
  • An action-focused review where subscription choices produce potential-savings totals. Its local HTML copy keeps every figure traceable to normalized rows and source filenames.
  • Secret Mission: Add a What Changed card for the largest category increase. The card links to its supporting transactions.

Are there any prerequisites?

You need a paid Claude subscription plus a Mac that supports Claude Desktop. You also need an internet connection.

The included money-dashboard.plugin file lets you follow the recommended path without coding experience.

Before We Start

A clear purpose keeps your dashboard focused on a real person. It also connects the dashboard to one monthly financial decision.

For example, the decision could involve reviewing subscriptions or choosing a savings transfer. Naming one decision gives every later dashboard view a practical goal.

Set Up Claude Desktop and a Safe Workspace

Your dashboard purpose gives this workspace a clear job. Local finance files need a controlled boundary before you import anything.

Claude Desktop lets Cowork work only inside folders you connect. You'll create that boundary before any synthetic data enters the project.

In this step, get ready to:
  • Prepare Claude Desktop for Cowork.
  • Set the privacy and approval boundaries.
  • Connect the finance-dashboard-lab folder to a Cowork project.
Get Claude Desktop ready

Cowork uses Claude Desktop to reach local files on your Mac. An active internet connection keeps that session connected.

I already have Claude Desktop

  • Press Cmd+Space to open Spotlight.
  • Type Claude and press Enter.
  • Sign in with your paid Claude subscription.
  • Select Cowork in the message box.

You'll see Cowork become the active workspace option.

Cowork Missing?

  • Visit the Claude Desktop download page.
  • Select macOS.
  • Select Download.
  • Open the downloaded installer.
  • Restart Claude Desktop through Spotlight after the installer finishes.
  • Look for Cowork in the message box again.

Still missing Cowork? Help me check my Claude Desktop setup.

I need Claude Desktop

The installer places Claude Desktop on your Mac.

  • Press Cmd+Space to open Spotlight.
  • Type Claude and press Enter.
  • Sign in with your paid Claude subscription.
  • Select Cowork in the message box.

You'll see Cowork become the active workspace option.

Installation Not Working?

Confirm that your Mac has an active internet connection. Run the downloaded installer again if Claude does not appear in Spotlight.

Need a hand? Help me troubleshoot Claude Desktop on my Mac.

That's the first gate cleared. Cowork is available in your signed-in desktop app.

Review privacy and permissions

Financial files deserve a clear privacy boundary. This project uses synthetic CSV files to keep sensitive identifiers out of the workflow.

Manual mode pauses before file actions so you can approve them. The privacy setting controls whether new chats may be used for future model improvement.

  • Open Settings in Claude Desktop.
  • Select Privacy.
  • Review Help Improve our AI models.
  • Turn off Help Improve our AI models if you do not want new chats used for future model improvement.
  • Switch back to Cowork from earlier.
  • Select Manual in Cowork's permission mode control.

You'll see Manual as the selected permission mode. File actions now pause for your approval.

How the Safety Boundary Works

  • Cowork can read or write only inside folders you connect.
  • Synthetic sample files are the only financial data used during this project.
  • A future real export must omit account numbers, card numbers, routing numbers, passwords, login credentials, and unnecessary identifiers.
Create or reconnect the project folder

The attached folder becomes Cowork's working boundary. Each empty subfolder reserves one type of finance data.

I need the project folder

  • Press Cmd+Space to open Spotlight.
  • Type Finder and press Enter.
  • Select Desktop in the Finder sidebar.
  • Select File in the menu bar.
  • Select New Folder.
  • Type finance-dashboard-lab and press Enter.
  • Open the new finance-dashboard-lab folder.
  • Create source/, config/, data/, and output/ inside it with Finder's New Folder command.

The Desktop now contains your empty project structure.

  • Return to Cowork from earlier.
  • Create a Cowork project using an existing folder.
  • Choose the finance-dashboard-lab folder on your Desktop.
  • Open the new Cowork project.

You'll see the local folder attached to the project.

I already have the project folder

  • Locate the existing finance-dashboard-lab folder in Finder.
  • Open the finance-dashboard-lab folder.
  • Check for source/, config/, data/, and output/.
  • Select File in Finder for the first missing subfolder.
  • Select New Folder.
  • Enter the missing folder name.
  • Repeat the same Finder action for each missing name.

The local folder now contains all four empty subfolders.

  • Return to Cowork from earlier.
  • Create a Cowork project using the existing folder.
  • Choose the finance-dashboard-lab folder when Cowork asks for a local folder.
  • Open the new Cowork project.

You'll see the local folder attached to the project.

What Each Folder Holds

  • The source/ folder is reserved for immutable CSV exports.
  • The config/ folder is reserved for approved preferences.
  • The data/ folder is reserved for normalized transactions and the import log.
  • The output/ folder is reserved for the generated dashboard.

Before the final check, consider whether Cowork should show all four folders with Manual mode selected. Hold that prediction.

  • Switch back to Claude Desktop from earlier.
  • Select Cowork in the message box.
  • Return to the open finance-dashboard-lab project.
  • Confirm that Manual is selected.
  • Expand the attached finance-dashboard-lab folder.

You'll see source/, config/, data/, and output/ inside the attached folder. You'll also see Manual selected as the Cowork permission mode.

Your safety foundation is in place. Cowork now has a controlled local workspace for the synthetic finance workflow.

Your safe workspace is ready. Next, you'll install the Money Dashboard plugin or build the same package yourself.

Install and Run Money Dashboard

Your finance-dashboard-lab workspace now gives Claude Cowork a controlled place to process synthetic CSV data. The dashboard still needs a reusable workflow that can turn those transactions into reviewed outputs.

Choose the supplied Money Dashboard plugin for the shorter route. Choose Plugin Create if you want to learn how the same package is assembled.

Why Are There Two Paths?

Both paths produce the same locally installed plugin. The Fast Track prioritizes using the finance workflow.

The Builder Track exposes the package structure through supplied source files. Your choice changes the setup time while preserving the final dashboard.

In this step, get ready to:
  • Install the supplied plugin or recreate it from the supplied source files.
  • Verify the plugin's skills and safety boundaries.
  • Generate an auditable January dashboard with Setup.
Choose your plugin path

The installation paths converge on the same package. Pick the route that matches how much of the plugin structure you want to inspect.

Fast Track, use the ready-made plugin

The ready-made package contains the supplied manifest, skills, references, and sample data. This route gets you to the finance workflow sooner.

  • Download the money-dashboard.plugin package from the shared Google Drive folder.
  • Select Customize in the current Cowork project.
  • Select Plugins.
  • Select Add.
  • Choose the downloaded money-dashboard.plugin package.

You'll see Money Dashboard added to your local plugin list.

Builder Track, create the plugin yourself

The Builder Track uses the supplied Complete Code files as the source of truth. Plugin Create assembles those files into the same local package used by the Fast Track.

  • Select Customize in the current Cowork project.
  • Select Plugins.
  • Select Browse plugins.
  • Find Plugin Create.
  • Select Install.

You'll see Plugin Create become available in Cowork.

  • Ask Plugin Create to build a skills-only plugin named money-dashboard from the supplied Complete Code files while preserving every path and filename.
  • Compare the generated package with every supplied Complete Code file.
  • Select Save plugin.

Builder Package Checkpoints

  • The .claude-plugin/plugin.json manifest identifies money-dashboard as Money Dashboard at version 1.0.0.
  • The skills/setup/SKILL.md, skills/customize/SKILL.md, and skills/monthly-update/SKILL.md files define the three workflows.
  • The references/dashboard-spec.md and references/preferences-template.json files define dashboard behavior and starting preferences.
  • The package includes sample-data/transactions-january.csv and sample-data/transactions-february.csv.
Verify the installed package

A plugin can bundle several types of capabilities. This project uses only reusable Skills plus local reference files and synthetic sample data.

  • Return to Customize in Cowork.
  • Select Plugins.
  • Select Money Dashboard.
  • Confirm the component list shows Setup, Customize, and Monthly Update.
  • Inspect the Setup skill.
  • Confirm the Setup skill treats files in source/ as read-only evidence.

What Each Skill Controls

Setup creates the first dashboard after an intake review. Customize proposes preference changes before updating normalized data.

Monthly Update imports later files with duplicate protection. Each skill preserves source CSVs so the dashboard remains auditable.

  • Inspect the Customize skill.
  • Inspect the Monthly Update skill.
  • Confirm the package includes transactions-january.csv.
  • Confirm the package includes transactions-february.csv.
  • Confirm the package contains no connector, MCP server, hook, external package, bank login, or API key.
  • Enable the plugin after the component review is complete.

Your Money Dashboard plugin is now enabled with three review-first skills. Its work stays inside the attached project folder.

Can't See All Three Skills?

  • Return to the Money Dashboard entry under Customize and Plugins.
  • Re-add money-dashboard.plugin if the Fast Track package did not finish installing.
  • Compare the Builder Track package paths with the supplied Complete Code files if a skill is missing.

Still missing a component. Help me verify my Money Dashboard plugin package.

Generate the January dashboard

Setup converts the January sample into normalized records, preferences, an import log, a local HTML dashboard, and a Cowork artifact. Manual mode keeps the file writes behind an approval checkpoint.

Before you start Setup, make a quick prediction about whether every merchant will receive an approved category automatically.

  • Select the Setup skill in Money Dashboard.
  • Choose sample mode.
  • Confirm transactions-january.csv as the intended source file.
  • Confirm USD as the currency.
  • Keep the savings target at 20 percent.
  • Confirm the supplied category list.
  • Confirm the starter category budgets.

Why Approval Comes Before Writes

The intake summary lets you verify the source file and row count before Cowork changes the attached folder. It also shows exactly where each generated output goes.

This checkpoint keeps the imported evidence separate from generated files. The source CSV remains unchanged after the import.

  • Review the intake summary.
  • Confirm the summary reports 10 January rows.
  • Confirm the planned outputs include config/preferences.json.
  • Confirm the planned outputs include data/transactions.json and data/import-log.json.
  • Confirm the planned outputs include output/finance-dashboard.html and the Money Dashboard artifact.
  • Approve the requested writes after every path matches the summary.

Setup copies the immutable January sample into source/. It creates the approved preferences and normalized outputs in their planned folders.

  • Confirm the completion report lists source/transactions-january.csv.
  • Confirm the completion report lists config/preferences.json.
  • Confirm the completion report lists data/transactions.json and data/import-log.json.
  • Confirm the completion report lists output/finance-dashboard.html.
  • Open the Money Dashboard artifact from the Setup result.
  • Select January in the dashboard.
  • Inspect the Needs Review rows.

You'll see MKTPLACE 4831 and MORNING BEAN in Needs Review. That result confirms the workflow preserves uncertainty for human review instead of silently approving ambiguous merchants.

You now have an interactive January dashboard plus a self-contained local copy backed by normalized rows and an import log. Every displayed figure has an auditable path back to the synthetic source file.

Your reusable finance workflow is running with its first reviewed month. Next, you'll turn the two unresolved merchants into category rules that future imports can reuse.

Save Reusable Category Rules

The January dashboard is now built from normalized sample transactions. Its Needs Review queue exposes two merchant descriptions that need your judgment.

Now you'll use the Customize skill in the Money Dashboard plugin to save that judgment as reusable preferences. Cowork remains in Manual mode so each write waits for your approval.

In this step, get ready to:
  • Propose reusable category rules with Customize.
  • Review every preference change before approval.
  • Audit the corrected January dashboard.
Propose the preference changes

Merchant descriptions often stay identical across monthly exports. A saved merchant rule lets the same approved category flow into future imports.

January contains one MKTPLACE 4831 row. It contains two MORNING BEAN rows.

  • In the Cowork project from earlier, select Customize from the enabled Money Dashboard plugin.
  • Ask Customize to map MKTPLACE 4831 to Shopping.
  • Ask Customize to map MORNING BEAN to Coffee.
  • Ask Customize to retain the 20 percent savings target.
  • Ask Customize to set the Coffee monthly budget to 30 USD.

Customize should present a before-and-after summary with the affected row counts. No project file should change before approval.

After you approve the proposal, Manual mode may pause on the files Customize needs to regenerate. Those prompts give you one more chance to protect the attached folder.

  • Compare the proposed before-and-after table with both requested category mappings.
  • Check that the affected-row summary counts one MKTPLACE 4831 transaction.
  • Check that the affected-row summary counts two MORNING BEAN transactions.
  • Confirm that forceReviewMerchants will contain neither merchant.
  • Confirm that the savings target remains 20 percent.
  • Confirm that the Coffee monthly budget becomes 30 USD.
  • Give Cowork approval for the proposed preference change.
Regenerate and audit January

Once approved, Customize reapplies the rules to the normalized January rows. Original transaction facts stay unchanged.

The source CSV remains immutable evidence. Customize changes the approved preferences plus the generated outputs.

The local HTML dashboard is generated from the corrected normalized rows. The Cowork artifact uses the same corrected data.

  • Approve the requested local writes only if each path sits inside finance-dashboard-lab.
  • Approve the Money Dashboard artifact update.

Customize should finish with a summary naming the changed files. It should also report the affected row count plus the unresolved review count.

  • Confirm that the changed local files include config/preferences.json, data/transactions.json, and output/finance-dashboard.html.
  • Confirm that the changed-file list omits source/transactions-january.csv.
  • Confirm that the changed-file list omits the installed plugin skill files.

Before you reopen the dashboard, do you expect any January row to remain in Needs Review after both approved rules are reapplied?

  • Reopen the Money Dashboard artifact from Cowork's Artifacts view.
  • Use the month selector to choose January.
  • Open the Overview view.

You should see a Needs Review count of 0.

  • Open the Transactions view.
  • Use the category filter to select Coffee.

You should see both January MORNING BEAN transactions categorized as Coffee.

  • Use the category filter to select Shopping.

You should see the January MKTPLACE 4831 transaction categorized as Shopping.

  • Return to Finder.
  • Open the finance-dashboard-lab/output folder.
  • Double-click finance-dashboard.html.

The local dashboard should show the same corrected January categories. Its Needs Review count should also be 0.

  • Return to Cowork.
  • Ask Cowork to display config/preferences.json without editing it.

What should the preferences show?

The categoryRules array should include MKTPLACE 4831 mapped to Shopping plus MORNING BEAN mapped to Coffee.

The forceReviewMerchants array should contain neither merchant.

The savingsTargetPercent value should remain 20.

The budgets.Coffee value should be 30.

🙋‍♀️ Still Seeing Needs Review?

  • Check that you approved both merchant mappings in the before-and-after summary.
  • Run Customize again if forceReviewMerchants still contains either merchant.
  • Open the regenerated file from finance-dashboard-lab/output if the local dashboard shows an older result.

Still stuck? Help me audit why my approved Money Dashboard category rules did not clear Needs Review.

Your January dashboard now carries your approved Coffee and Shopping rules with zero unresolved rows. Next, you'll add February while Monthly Update protects the totals from a duplicate transaction.

Import February Without Double-Counting

Your January dashboard is clean because its merchant rules now reflect your decisions. A new CSV export can overlap with data you already imported.

The Monthly Update skill checks February before any rows join the dashboard. You will review its import summary before approving the merge.

In this step, get ready to:
  • Copy the February sample into the attached project folder.
  • Review the Monthly Update import proposal.
  • Verify the refreshed dashboard across both months.
Add the February sample

Source CSVs are immutable evidence. The February file stays unchanged so every dashboard value remains traceable to an original row.

  • Switch back to the finance-dashboard-lab project in Claude Cowork.
  • Ask Claude to copy the bundled transactions-february.csv file into source/.
  • Review the Manual mode approval request for the copy.
  • Confirm the destination is finance-dashboard-lab/source/transactions-february.csv.
  • Approve the file copy.
  • Ask Claude to confirm the files now stored in source/.

You should see transactions-january.csv plus transactions-february.csv in the source folder.

That copy gives the updater a second source file while preserving the bundled sample.

Before you run Monthly Update, do you expect every source row to become a new February transaction?

  • Select the Monthly Update skill from the installed Money Dashboard plugin.
  • Choose transactions-february.csv when Claude lists files that have not been imported.
  • Review the detected filename.
  • Review the detected date range.
  • Review the source row count.

You should see transactions-february.csv with 11 rows. Its date range starts on 2026-01-18 because one row reaches back into January.

That January date exposes the intended overlap. The import needs review before those rows can join your totals.

February file not detected?

  • Confirm transactions-february.csv sits directly inside finance-dashboard-lab/source/.
  • Confirm the February file was not placed inside the bundled sample-data/ folder.

Still missing the file? Help me find why Monthly Update cannot detect the February CSV.

Review duplicate handling before the merge

A non-empty transaction_id provides strong duplicate evidence. An exact match means the transaction already exists in the normalized store.

  • Review the proposed count of new rows.
  • Confirm the exact transaction-ID duplicate count is 1.
  • Inspect the exact duplicate detail for tx-007.
  • Check the potential duplicate count.
  • Review every fallback potential duplicate manually if the count is above 0.
  • Check the invalid row count.
  • Check the Needs Review count.

The proposal should show 10 new rows plus 1 exact duplicate. It should show 0 potential duplicates.

You should also see 0 invalid rows plus 0 Needs Review rows.

Why is tx-007 excluded?

  • The February-file tx-007 has a non-empty transaction_id that exactly matches the imported January transaction.
  • The matching row describes the same MORNING BEAN purchase from 2026-01-18.
  • The fallback dedupe_key check is unnecessary for this row because its transaction ID is reliable.
  • The source CSV remains unchanged as audit evidence.
  • Confirm the proposal keeps tx-007 out of the merge.
  • Approve the merge after every summary category has a separate count.

That safeguard just did its job. February joins the dashboard without counting the repeated coffee purchase twice.

The approved merge adds the new rows to data/transactions.json. The decision is recorded in data/import-log.json.

Compare the two months

The refreshed views should tell the same two-month story. February should reuse the merchant rules you approved for January.

Before you reopen the dashboard, which saved categories do you expect February's ambiguous merchants to use?

  • Reopen the Money Dashboard artifact from the Cowork project.
  • Select January from the month selector.
  • Select February from the month selector.
  • Open Trends.
  • Confirm MORNING BEAN appears under Coffee for February.
  • Confirm MKTPLACE 4831 appears under Shopping for February.

You should see January plus February in the dashboard. The saved Coffee and Shopping rules now work across both months.

A recurring candidate appears when the same normalized merchant has supporting charges in at least two different months. Its status remains informational until you record a decision.

  • Open Actions.
  • Confirm STREAMBOX appears as a recurring candidate.
  • Inspect the January supporting row for STREAMBOX.
  • Inspect the February supporting row for STREAMBOX.
  • Confirm CLOUDSTORE appears as a recurring candidate.
  • Inspect the January supporting row for CLOUDSTORE.
  • Inspect the February supporting row for CLOUDSTORE.

Both candidates should show evidence from January plus February. This proves the dashboard uses transaction history before suggesting a review.

The Cowork artifact and local HTML dashboard use the same normalized data. Checking the local copy confirms the portable output was regenerated too.

  • Switch back to Finder.
  • Navigate to finance-dashboard-lab/output/.
  • Double-click finance-dashboard.html to open it in Safari.
  • Select February from the local dashboard's month selector.
  • Open Transactions.
  • Confirm transactions-february.csv appears as a source filename.
  • Return to finance-dashboard-lab/data/ in Finder.
  • Double-click import-log.json to inspect the February entry.
  • Confirm the entry records transactions-february.csv as the source file.
  • Confirm the entry records one excluded exact duplicate.
  • Confirm the excluded transaction is tx-007.
  • Confirm the entry records your approval decision.

Your import log now explains exactly how February entered the dashboard. The visible totals remain traceable to approved normalized rows.

Do the months or candidates look wrong?

  • Check that the February import was approved before reopening the dashboard.
  • Confirm tx-007 appears in the import log as excluded.
  • Confirm both source filenames appear in the transaction views.

Need help tracing the mismatch? Help me audit my two-month Money Dashboard outputs.

Your dashboard now handles overlapping exports without inflating its totals. Next up, you will turn those recurring candidates into recorded monthly decisions.

Record Monthly Review Decisions

Your Money Dashboard now compares January with February. The repeated tx-007 appears only once.

Recurring charts become useful when they lead to a recorded choice. The Customize skill saves those choices for each monthly review.

In this step, get ready to:
  • Review the transactions supporting each recurring candidate.
  • Record subscription decisions through the Customize skill.
  • Verify the refreshed Actions view against your monthly plan.
Record subscription decisions

A recurring candidate becomes auditable when each decision has supporting rows. The saved status then controls whether the charge contributes to potential savings.

What do the statuses mean?

  • The Could Cancel status records a charge you may remove. It contributes to potential savings.
  • The Keep status records an intentional subscription. It stays out of potential savings.
  • Supporting rows show the months behind each recurring candidate. Their source filenames preserve the audit trail.
  • In the Money Dashboard plugin, select Customize to start the preference review.
  • Request STREAMBOX status Could Cancel.
  • Request CLOUDSTORE status Keep.
  • Review the proposed before-and-after table.
  • Inspect the supporting transactions for STREAMBOX.
  • Inspect the supporting transactions for CLOUDSTORE.

You should see January rows for each merchant. You should also see February rows for each merchant.

The source filename beside each row connects the proposed decision to imported evidence.

The next approval updates config/preferences.json. Manual mode keeps the write paused until you choose to continue.

  • Approve the proposed preference update.
  • Review the completion report for the updated config/preferences.json file.
  • Confirm the report lists the regenerated output/finance-dashboard.html file.
  • Confirm the report lists the refreshed Money Dashboard artifact.

That review is now reusable in future months. STREAMBOX is saved as Could Cancel.

CLOUDSTORE is saved as Keep. The normalized transaction facts remain unchanged.

Decisions missing after approval?

  • Return to the completed Customize task.
  • Check whether a write approval is still waiting in Manual mode.
  • Run the Customize skill again if the completion report does not list the updated preferences.

Still stuck? Help me check why my subscription decisions were not saved.

Complete the monthly review

Potential savings uses recorded subscription decisions. This keeps the calculation tied to choices you approved.

How does potential savings work?

  • Potential monthly savings totals the latest recurring charges marked Could Cancel or Cancelled.
  • Potential annual savings multiplies the monthly total by 12.
  • Charges marked Keep remain visible. They are excluded from both savings totals.

Before you reopen Actions, which saved status do you expect to add to potential savings? The check below tests your prediction.

  • In Cowork, reopen the Money Dashboard artifact.
  • Select Actions.

You should see STREAMBOX marked Could Cancel. CLOUDSTORE appears as Keep.

Potential monthly savings shows 15.99 USD. Potential annual savings shows 191.88 USD.

You have turned the subscription review into a reusable monthly decision.

The plugin regenerates its artifact from the same normalized data used by the local dashboard. Checking both surfaces confirms that the saved decisions reached each output.

  • In Finder, navigate to finance-dashboard-lab/output.
  • Double-click finance-dashboard.html to open the refreshed local dashboard in Safari.
  • Select Actions in the local dashboard.

You should see the same 15.99 USD monthly figure. The annual figure should remain 191.88 USD.

  • Inspect the recurring support rows for their source filenames.
  • Compare the latest actual Savings Transfer with the suggested savings-transfer target.
  • Confirm the dashboard labels its generated observations as informational.
  • Confirm the dashboard states that its observations are not financial advice.

You should see transactions-january.csv beside January evidence. You should see transactions-february.csv beside February evidence.

The dashboard shows the actual Savings Transfer beside the user-defined 20 percent suggested target. Every decision remains connected to its supporting transactions.

Secret mission

Add a What Changed Card

Your dashboard already shows monthly totals. Extend it with a What Changed card that identifies the largest category increase, shows both month values, and links the result to supporting transactions.

Clean Up Your Resources

Clean Up Your Resources

Your existing paid Claude subscription covers this project with no additional project charge. Decide whether to keep the saved workflow, pause your work, or remove its local resources.

Resources you used:

  • The locally installed Money Dashboard plugin with the Setup, Customize, and Monthly Update skills.
  • The customized references/dashboard-spec.md presentation requirements for the What Changed card.
  • The finance-dashboard-lab folder with synthetic CSVs plus config/preferences.json, data/transactions.json, data/import-log.json, and output/finance-dashboard.html.
  • The related Claude Cowork task containing your workflow history.

Keep everything running

No action is needed. Choose this option if you want to reuse the dashboard for future monthly reviews.

  • Leave the Money Dashboard plugin enabled.
  • Keep finance-dashboard-lab attached to Cowork.
  • Run Monthly Update with a redacted CSV during your next monthly review.
  • Keep the related Cowork task as a record of your workflow.

Pause - I'll come back to this later

No cloud or local infrastructure remains running after Claude Desktop closes. Your plugin plus project files stay saved.

  • Close Claude Desktop.
  • Retain finance-dashboard-lab in its current location.
  • Leave the Money Dashboard plugin installed.
  • Keep the related Cowork task for your next review.

Delete - I don't want to use this again

This option permanently removes the local plugin plus the project folder. Your paid Claude subscription stays unchanged.

  • Copy finance-dashboard-lab to a separate backup location if you want to preserve its history.
  • In Claude Desktop, select Customize.
  • Select Plugins.
  • Select Money Dashboard from the installed plugins.
  • Select Uninstall.

Good work. The reusable Money Dashboard plugin is now removed from this Mac.

  • Switch to Finder.
  • Locate the attached finance-dashboard-lab folder.
  • Drag finance-dashboard-lab to the Trash.
  • Empty the Trash after checking that your backup is safe.
  • Return to Cowork.
  • Open task history.
  • Select the related Money Dashboard task.
  • Select Delete.

Cleanup is complete. You should no longer see Money Dashboard in the installed plugins list.

The finance-dashboard-lab folder should be absent from Finder. The related task should be absent from Cowork task history.

Archiving a Cowork project alone does not delete its attached local folder.

Nice Work!

Nice Work!

You made it! Your Money Dashboard plugin now turns monthly transaction CSVs into an auditable financial review inside Claude Cowork.

You've learned how to:

  • Create a repeatable finance workflow with the Setup, Customize, and Monthly Update skills. Use that workflow to produce an interactive Cowork artifact plus a self-contained local HTML dashboard.
  • Turn human-reviewed corrections into reusable merchant rules. Save budgets, a savings target, and subscription decisions in config/preferences.json.
  • Apply duplicate protection by excluding exact duplicate tx-007. Preserve an audit trail through source filenames, data/transactions.json, and data/import-log.json.
  • Complete the optional Secret Mission by adding a What Changed card that identifies the largest category increase. Link the card to the supporting transactions without counting transfers.

Ready to quiz yourself?