On this page
- What kinds of Salesforce org cleanup tools exist, and what does each one do?
- What should I check before I trust a tool with my org?
- How do I find unused fields in Salesforce without a tool?
- How do I decide which fields, flows and permission sets are safe to remove?
- How do I run a cleanup that I can undo?
- What does a cleanup tool cost compared with doing it by hand?
- When is a free scan enough, and when do I need a tool that makes changes?
- Sources
- Common questions about Salesforce org cleanup tools
A Salesforce org cleanup tool is a tool that scans your org's metadata for unused fields, flows and permission sets, and ranks what is safe to remove. Free field analyzers cover fields only, and a tool that changes your org also needs a sandbox and your approval.
Clientell AI, the AI Salesforce admin, scans your org read-only and returns a ranked clean-up list with the change it would make for each item. Start with the free technical debt audit. For empty and duplicate data, see the Clientell AI data cleaning tool. Clientell AI makes one of the tools compared here.
What kinds of Salesforce org cleanup tools exist, and what does each one do?
Four kinds exist. The table uses what each vendor says on its own site, so check the current version before you buy.
Swipe sideways to see all columns
| Kind | Example | What it gives you | What it leaves to you |
|---|---|---|---|
| Field-usage analyzer | FieldRat, FieldSpy by Arovy | Fill rate and references for each field. FieldRat scores deletion risk from 0 to 100. Both list a free offer. | You delete the field and fix what breaks. Fields only. |
| Native documentation app | Technicaldebts.com | Lists unused custom objects and fields, and documents what each item is for. | You decide and act. The vendor's consultants check its AI-written documentation, per its site. |
| Assessment and ranked list | Kicksights | A ranked fix list built from org health signals. | You or their team make the changes. |
| AI admin that scans and changes | Clientell AI | A graded read-only scan with a 24-hour fix plan, then the change. | Clientell AI asks before it changes anything in production. |
FieldRat runs a Quick Scan over 11 metadata types and a Deep Scan over 23, and says it exports schema, not record values. It is the better pick when your whole job is retiring fields and you want a 0 to 100 risk score on each one.
FieldSpy says it finds unused fields, shows how often each field is populated, and flags references to automation. It is a good free start for a single object.
Technicaldebts.com says it finds custom objects with 0 records and no reference in Apex or elsewhere, and that a Full or Partial Copy sandbox is enough to scan. It fits best when you also need documentation of what the org does.
Kicksights groups findings by risk, effort and owner, and offers help to scope and build the fix. It suits a team that wants a ranked to-do list and outside hands to do the work.
Clientell AI's technical debt audit page describes a free, read-only scan with A to F grades across 6 domains (Apex classes, Apex triggers, flows, Lightning Web Components (LWC), workflow rules and metadata) and a 24-hour fix plan. It is not a field-only analyzer, so a dedicated field tool may go deeper on one object.
What should I check before I trust a tool with my org?
Run any tool through this list before you connect it:
- Is the scan read-only? Ask which permissions it requests and whether it installs a package.
- Does it export record values, or only schema? FieldRat says schema only. Others may differ.
- How deep is the dependency check? A field can sit in a flow, a report, a layout and an integration. Count how many metadata types the tool reads.
- Can you scan a sandbox first? Technicaldebts.com says a Full or Partial Copy sandbox is enough. Salesforce explains the sandbox types in its help.
- Does it show each change before it runs, wait for approval, and keep a way back?
- What evidence does it give per item: last use, fill rate, references?
- Who owns the result: your team, or the vendor's consultants?
Treat every scan as a briefing, not permission to delete. A metadata review has limits, because it cannot see what happens outside Salesforce.
You can also cross-check any tool's dependency list yourself. Salesforce exposes the MetadataComponentDependency object in the Tooling API, and you can query it with SOQL (Salesforce Object Query Language) for the components that reference one field. If the tool misses a reference that this query shows, do not trust its "safe to delete" label.
How do I find unused fields in Salesforce without a tool?
You can check one field by hand in about ten minutes. Use this when the list is short or you want to test a tool's answer.
- Open Setup, then Object Manager, then the object (for example Account), then Fields & Relationships, and click the field. Click Where is this used? to list the layouts, flows, reports and Apex that reference it.
- In Developer Console, open Query Editor and tick Use Tooling API. Run
SELECT MetadataComponentName, MetadataComponentType FROM MetadataComponentDependency WHERE RefMetadataComponentName = 'Legacy_Region__c'. Each row is one component that uses the field. - Count the filled records with
SELECT COUNT() FROM Account WHERE NOT Legacy_Region__c = null. A count of 0 means nobody has filled it. - Check recent writes with
SELECT MAX(LastModifiedDate) FROM Account WHERE NOT Legacy_Region__c = null. An old date means no one has edited a record that holds a value, though other fields may have changed it.
Expected result: a field with 0 dependency rows, a count of 0 and no recent date is a safe candidate. Failure case: step 2 returns nothing for a field that an outside system writes to, because the query only sees Salesforce metadata. Undo: nothing changed yet, since every step above only reads.
How do I decide which fields, flows and permission sets are safe to remove?
Remove an item only when three signals agree: no recent use, no references, and no owner who objects. Salesforce's own Trailhead module on cleaning up unused fields teaches the same habit of checking a field's use before you retire it.
Swipe sideways to see all columns
| Item | Evidence you need | Red flag |
|---|---|---|
| Custom field | Fill rate, last write date, and no reference in flows, Apex, reports or layouts | An integration writes to it. |
| Flow | An inactive or draft version, and no subflow or Apex call | A schedule or another flow calls it. |
| Permission set | No assigned users, and no group membership | An integration user holds it. |
The red flags come from the usual cleanup failures. A cleanup removes a field that an integration quietly writes to. A permission change removes access a small team relied on. An automation change collides with a validation rule nobody remembered.
For example, Account.Legacy_Region__c has a very low fill rate, so it looks safe to delete. A monthly billing integration fills it at month end, and the fill rate misses that. The last write date catches it.
How do I run a cleanup that I can undo?
Run it as a series of small, reversible steps:
- Export the items you plan to remove, with their metadata, so you have a copy. Retrieve the fields with the Metadata API, and export the data with Data Export under Setup.
- Deactivate or hide each item first. Deactivate a flow in Flow Builder. Remove a field from page layouts and from field-level security (FLS, the setting that controls who can see a field).
- Wait through one full business cycle, such as a month-end close, to catch the integrations that only run then.
- Delete a small batch in a sandbox, then in production, and check the same reports each time. For metadata, Salesforce documents deleting components with a destructive changes file.
- Keep the export until the next release cycle passes with no complaints.
Expected result: after step 3, nothing in your reports or integration logs changes. If a report breaks, restore the field to the layout from your export and the deactivation is undone.
A tool earns its price at step 4. Clientell AI offers a read-only scan, sandbox testing, a plan for each change, and rollback. It asks before it changes anything in production. See security.
What does a cleanup tool cost compared with doing it by hand?
Doing it by hand costs admin hours. Tracing one object's fields through flows, reports, layouts and Apex can take a full working day, and admin pay data is in the Salesforce Ben careers guide.
The free analyzers above cost nothing to try. Clientell AI's audits are free and read-only. Clientell MCP is $249 a month and Clientell Platform is $500 a month, which adds deployments and weekly audits. See pricing, the source of truth.
When is a free scan enough, and when do I need a tool that makes changes?
Swipe sideways to see all columns
| If this is true | Then |
|---|---|
| You will make the changes yourself within one sprint | A free scan is enough. |
| The list is mostly fields and reports | A free analyzer is enough. |
| Flows and permission sets are on the list | You need dependency checks and tested changes. |
| The backlog is longer than your time | You need a tool that makes the changes. |
| An auditor will ask who approved what | You need a record of each approval. |
Try it in Clientell AI: "Scan my org in read-only mode and list the custom fields nothing uses, with the reports, flows and page layouts that would break if I deleted each one."
Run the free, read-only technical debt audit, or start the 14-day trial, no card. For the bigger picture, read the technical debt guide.
Common questions about Salesforce org cleanup tools
What is the best free Salesforce org cleanup tool?
It depends on the job. FieldRat and FieldSpy list free offers for field analysis. Clientell AI's technical debt audit is a free, read-only scan across six code domains. Compare what each reads before you choose.
Is it safe to delete unused fields?
Only with the evidence in the table above. A scan cannot see an outside system that writes to a field, so deactivate first.
Can an AI agent clean up my org?
It can, if it reads your org first, shows each change, and asks before it changes anything in production. Without those, treat it as a risk.
Does a cleanup tool need a sandbox?
Use one for the first run. A sandbox shows what breaks before production does. Technicaldebts.com says a Full or Partial Copy sandbox is enough for its scan.
How often should I run a cleanup?
Run a scan before each major release and before any Agentforce rollout. Delete in small batches so each change stays easy to undo.
Sources
Salesforce documentation
- MetadataComponentDependency (Tooling API)
- Deleting components with destructive changes (Metadata API)
- Clean up unused fields (Trailhead)
- Sandboxes (Salesforce Help)
Industry research
Last reviewed October 2026
Last updated: October 22, 2026



