Changelog

Patchy's system view was talking to itself. The panel that shows what Patchy has figured out about your system was full of internal wording — grader, claims, audit sample, consent model — that meant something to the machine and nothing to a person skimming it. This release rewrites that surface in plain language and makes the questions Patchy asks actually useful: each one now names the module it's about, shows the specific things that no longer line up, and gives you answer buttons that match the question instead of a bare Yes/No.

It also closes a gap where answering did nothing. When Patchy asks whether a module is still a live part of your system and you tell it that one is test or leftover, it now clears that module's flagged mismatches so they stop cluttering your review — instead of quietly recording your answer and changing nothing. Nothing in your GHL account is touched; this only affects what Patchy keeps asking you about.

The asset table's Status column also got smarter. Instead of echoing the raw stored status, it now shows an at-a-glance verdict — Ready, Active, Needs review, Has issues, Deprecated — and clicking it opens a short breakdown of why, so you can tell which assets actually need attention without opening each one.

A cluster of import-completeness fixes, all found by watching real GHL accounts import less than they should. Forms tucked inside folders were being skipped — the forms list only returned the first ten, so any account with more than that (or forms nested in a folder) imported a partial set, which quietly broke the workflow trigger links that pointed at the missing forms. The same silent-truncation problem hit surveys, funnels, and voice AI agents on larger accounts. All of them now page through the full list.

Token-based imports got an important fix too. Since late April, PatchyHub had been sending browser-session headers on token requests, and GHL started rejecting those as a permission error — so valid token imports were being misreported as "missing scope." Token imports now use the right headers and work again. And an account that simply doesn't have Custom Objects turned on no longer shows up as a phantom import failure.

Link accuracy improved as well: Patchy is better at matching workflow references to the right custom fields and values, and it stops flagging built-in GHL attributes as if they were your own custom fields. Existing accounts heal on the next import — nothing to redo.

Browser Connect imports now give you better evidence when something goes wrong. Instead of staring at a vague import failure and guessing whether GHL, the browser extension, or PatchyHub lost the plot, the import flow can now collect diagnostics you can download and send over for support.

The connector also does a stricter check that the captured GHL session is real before starting an import. That should cut down on the deeply annoying situation where everything looks connected until the first real request quietly fails.

Patchy's system view now has the first account health findings. These are the practical little red flags that should not be buried in raw asset lists: status mismatches, suspicious tags, and other account-level issues that deserve a human decision.

This release also tightened the brain's understanding of workflow links. Trigger-based references are resolved more accurately, unresolved references are preserved for later review, and cross-system lookups are less likely to attach evidence to the wrong place. The old AI Deep Dive report has been retired; the release readiness checklist remains.

PatchyHub now has a proper public guide library for the problems people are actually searching: cleaning up messy GHL accounts, documenting client onboarding, updating snapshots safely, onboarding VAs, choosing outsourced support, and understanding where AI documentation fits.

Inside the app, Patchy's understanding work got a stronger review loop. It can apply verdicts to live claims, ask better human questions when it needs context, and use a clearer understanding score. Translation: fewer stale claims hanging around pretending to be useful.

Patchy chat got a model upgrade and a round of answer discipline. It should handle larger answers with more breathing room, be less sloppy with broad asset searches, and format Closebot readiness answers more consistently.

The trust controls also got stricter. Patchy should be clearer about what it can see, when workflow internals are available, and when it needs to avoid pretending it inspected something it did not inspect. Humanity continues its long war against confident nonsense.

The pricing language now matches the product: AI is unlimited on paid plans, with sensible safety limits in the background instead of user-facing credit math. Agency AI needs also have a clearer contact path instead of being jammed into the normal plan comparison.

A few quality-of-life fixes landed in the same window. The setup wizard no longer drops people into an empty dead end, and GHL fields are easier to link and recognize across related objects.

PatchyHub's GHL merge field reference was rebuilt from GHL's current picker and documentation set. That gives PatchyHub a cleaner base for recognizing and explaining merge fields instead of relying on stale assumptions.

Patchy also got a small but important honesty pass. It should stop offering to pull proposal details from places where that tool is not available, and it should avoid promising an inspection it cannot actually perform.

The pricing page now reflects the simpler plan lineup: Free and Pro, without the old Founding Member and More Pro tiers hanging around like product archaeology. Free AI chat is described plainly, without credit jargon.

Patchy's understanding layer also became more visible and more actionable. Systems can show their own understanding items, Patchy can be manually asked to work on understanding, and background cadence keeps the work moving without requiring constant button pushing.

Workspace admins can now manage GHL fields, which matters because field cleanup is rarely a solo-owner-only job. A few rough edges around workspace state, module creation, and invites were cleaned up at the same time.

Patchy's background brain also got a reliability pass for large systems. The important user-facing bit: it should waste less time re-reading unchanged data, handle big module context more gracefully, and recover better when a background state check fails.