Approving Fixes¶
After Patcherly analyzes an error and generates a fix, you need to approve it before it's applied. This guide explains the approval workflow.
Two types of action in the workflow¶
Patcherly has two separate human steps in the error workflow. Knowing the difference makes the rest of this page (and auto-analysis vs auto-apply) much easier:
Starting analysis (Analyze with AI)¶
- When: the error is Pending (just detected).
- What it does: queues AI analysis. After you click, the error becomes Pending analysis while AI works on a fix suggestion.
- How: click Analyze with AI on a pending error in the Errors page.
- Auto-analysis skips this click: if your plan includes auto-analysis and it's turned on for the target, errors move straight to analysis. See Auto-Analysis and the Pricing page.
Patch approval (Approve patch)¶
- When: the error is Analyzed or Ready to Patch (AI has generated a patch).
- What it does: approves the suggested patch so Patcherly can apply it on your environment. The error becomes Approved, then Applying, then Patched if the apply succeeds.
- How: review the patch details and click "Approve" (described below).
- Auto-apply skips this step: when your plan includes the auto-apply feature and the per-target Auto Apply toggle is on (and the other conditions in the auto-apply section below are right), Patcherly approves the patch on your behalf and the connector applies it without your click.
- After approve: the status moves to Waiting for connector or Applying…. If Patcherly cannot reach your site, you will see a red Dispatch failed / Apply stalled badge — use Retry Patch. The Errors page legend (Show all) lists these badges by workflow stage.
These two steps are independent: clicking Analyze with AI does not approve the resulting patch. The patch has its own safety rail — a minimum confidence check, explained below.
Fix Approval Workflow¶
Step 1: Review the Fix¶
When a fix is ready for review:
- Navigate to Errors in your dashboard
- Find the error with status "Analyzed"
- Click on the error to view details
- Review the fix suggestion
Step 2: Understand the Fix¶
The fix details show:
- What will change: Files and code that will be modified
- Confidence score: AI's confidence in the fix (0-100%)
- Explanation: Why this fix was suggested
- Impact: What the fix will do
Step 3: Make a decision¶
Minimum confidence — the safety rail¶
Every fix comes with an AI confidence score from 0 to 100%. Your workspace has a minimum confidence set by Patcherly (default 90%) — a built-in guardrail that prevents low-confidence fixes from applying silently. If your plan includes auto-apply, the workspace owner can optionally raise that bar even higher for their own approvals on Profile → AI. The override only matters at the auto-apply gate, so it's hidden on plans that don't include auto-apply.
If a fix is below the threshold:
- The dashboard asks you to confirm before you approve or accept it (this also covers some bulk actions).
- Patcherly will not auto-approve a low-confidence fix on its own — you have to approve it manually in the dashboard.
- On live (non–dry-run) targets, you may get an email or in-app notice depending on your notification settings.
If a fix is at or above the threshold and your plan includes auto-apply with the Auto Apply toggle on for that target (and the other conditions below), Patcherly may auto-apply it (see the next section).
Custom scripts or external tooling — the dashboard's confirmation rules are the safety floor. Any other automation talking to your workspace must respect the same minimum-confidence guardrail.
Approve¶
Click "Approve" to:
- Queue the fix for application
- Fix will be applied automatically
- Backup created before application
- System monitors the result
Note: After a fix is applied, you may need to restart your application for the changes to take effect. See Restarting Your Application for details.
Reject patch¶
After AI analysis, click Reject patch to open a modal. You must pick an outcome before anything is saved — closing without a choice does nothing.
| Choice | Result |
|---|---|
| I patched it myself — used Patcherly suggestion | Error becomes Patched; counted as manual patch (suggestion) |
| I patched it myself — my own code | Error becomes Patched; counted as manual patch (own code) |
| I don't want or need to fix this | Error becomes Ignored; counted as ignored after reject |
For Pending errors you have not analyzed yet, use Hide Error & Ignore (hide, restorable) or Delete (remove the record; it may reappear if the bug still fires on the target).
Mark as manually fixed¶
On post-analysis errors, Mark as manually fixed is a separate two-choice modal (Patcherly suggestion vs own code). It does not chain through reject. Use it when you fixed the issue without applying the AI patch.
Hide Error & Ignore (pre-analysis and post-apply only)¶
Hide Error & Ignore is available on pre-analysis rows (Pending, Pending analysis, Analysis failed, Excluded) and on post-apply rows (Failed, Rolled back, Rollback failed) to tidy your list.
Click Hide Error & Ignore to:
- Mark the error as Ignored in the list (restorable via Unignore)
- Upsert an ignore pattern so the same signature does not auto-analyze again
- Pre-analysis hide does not increment Business Metrics totals; post-apply hide keeps your rollback/fix metrics unchanged
Not available after analysis when a fix proposal exists (Analyzed, Ready to Patch, Manual review required). Use Reject patch or Mark as manually fixed instead.
Auto-apply: when Patcherly applies a fix without your click¶
Auto-apply is when Patcherly approves and applies a fix on your behalf, without you clicking Approve the Fix in the dashboard. It's a separate paid feature with its own per-target toggle, independent from auto-analysis.
Auto-analysis vs auto-apply¶
| Auto-analysis | Auto-apply | |
|---|---|---|
| What you skip | The manual Analyze with AI click | The Approve patch step |
| How you turn it on | The Auto Analysis toggle, on each target (Profile → AI tab, or the Targets page row) | The Auto Apply toggle, on each target (Targets page row). Greyed with an upgrade hint when your plan doesn't include the feature. |
| Plan requirement | Paid plan that includes auto-analysis | Paid plan that includes auto-apply (Pro by default) |
The baseline path is always available on every plan: analyse → you approve the fix in the dashboard → the connector applies. Auto-apply is an optional layer on top, and it's now independent from auto-analysis at the feature level — a workspace can have auto-analysis ON and still keep auto-apply OFF if it wants every approve to go through the dashboard.
When auto-apply happens¶
Patcherly only applies a fix without your click when all four of these are true:
- Auto Apply is on for the target. The new per-target toggle (Targets page → expand row → Auto Apply). Requires your plan to include auto-apply.
- Auto Analysis is on for the target. Auto-apply only chains in after an auto-analysed fix; if auto-analysis is OFF, the connector pauses for your Analyze with AI click and won't reach the apply step automatically anyway.
- Dry-run is off for the target. Targets in dry-run still run the full automated pipeline, but the actual file write becomes a preview — nothing changes on disk. The dashboard surfaces a hint when Auto Apply is ON and dry-run is still ON so you know you'll only see previews.
- The fix's confidence is at or above your workspace minimum (default 90%). Below that, Patcherly always pauses for your manual approval — exactly like the baseline path. Connectors that get a
409 low_confidence_confirmation_requiredfrom the server stop quietly and the fix waits for review.
If any of the four is missing, the fix waits for your click in the dashboard, just as it does on a Personal plan. The server also re-checks that auto-apply is still on your plan and that the per-target toggle is still on at approve time — if either changed between detection and approve, the connector receives 409 auto_apply_not_enabled and stops, even if the detection response had said auto-apply was ON.
Where to set the controls¶
- Auto Analysis toggle: Profile → AI tab → Auto-Analysis Preferences, or expand a target row on the Targets page and use the Auto Analysis toggle. Both places control the same setting.
- Auto Apply toggle: Targets page → expand a target row → Auto Apply toggle. Greyed with an upgrade hint on plans that don't include the feature.
- Dry-run: edit the target on the Targets page. New targets are created with Dry-run on by default so a fresh workspace can't accidentally start writing files on day one.
- Workspace minimum confidence: Patcherly maintains a default for your workspace; contact support if you need it changed. The workspace owner can also set their own personal floor on Profile → AI when their plan includes auto-apply.
Bulk Approval¶
For multiple errors:
- Select errors using checkboxes
- Click "Bulk Approve" (or the low-confidence bulk confirmation when offered)
- Eligible fixes are approved; entries that fail policy (paths, confidence, state) are skipped or called out in the result
Same minimum confidence rules apply: the UI may ask you to confirm when any selected fix is below the threshold.
Approval Best Practices¶
- Review high-impact fixes: Always review fixes that modify critical code
- Check confidence scores: Higher confidence = more reliable fix
- Test in staging first: Apply fixes to staging before production
- Monitor after approval: Watch for any issues after fix is applied
Approval History¶
All approvals are logged:
- Who approved the fix
- When it was approved
- What was changed
- Result of the application
You can review approval history on the error details page, or on the Audit Logs page filtered by the matching Error ID to see every event tied to that error (analysis, approval, apply, rollback).
Next Steps¶
- Auto-Analysis — Skip the manual Analyze with AI click for faster error processing
- Restart your application after fixes
- Learn about rollback
- Apply manual fixes
- Understand error states
- Audit Logs — Per-event record of approvals, applies, and rollbacks