A '12 Silverado pulls into the bay with a transmission that's surging at low speed. The customer says it started a few weeks ago. You vaguely remember the truck — you think you flushed the trans on it last winter, or maybe that was the silver one. The shop management software shows the invoice for the flush. It doesn't show the conversation you had with the owner about hauling a fifth-wheel through the Rockies, or the note you scribbled on the inspection form about the rear main looking damp, or the fact that your apprentice mentioned the vehicle had been in twice for the same complaint and you'd asked him to keep an eye on it.
That's the gap. The shop management software handles invoicing and inventory. It rarely handles the working memory of the technicians. So most shops end up with a parallel notes system — usually a clipboard, sometimes a notebook per bay, occasionally a shared text document — that's exactly as findable as you'd expect.
This post is about replacing that parallel system with a notes workspace that the techs can actually use, and an assistant that can read across every vehicle in the shop's history in seconds.
What a working shop needs from a notes system
Before getting into the mechanics, the shape matters. A small auto shop needs a notes system that does four things:
- Per-vehicle diagnostic history — what was found, what was fixed, what was deferred, what the customer was told
- Cross-vehicle pattern recognition — every coolant system on a 2.7-liter EcoBoost we've seen, every Honda Pilot with a similar misfire code
- Customer context — how this customer talks about problems, what their tolerance for cost is, what they tend to defer
- Shop operations — supplier accounts, equipment maintenance, the running list of warranty claims, training notes for new hires
Plus it needs to be usable in the bay — meaning a tech can open a phone or tablet, find the vehicle, add a note, and get back to work without fighting the software. The shape is similar to what we describe in How to Use AI Notes for Fleet Management and AI Notes for Facilities Management — a vault of physical assets with running maintenance history.
A page per vehicle, organized by VIN
In Docapybara, every vehicle gets a markdown page. The simplest convention: the page title is the year/make/model and a short identifier. Inside the page, you keep VIN, customer name, and a running log — date, mileage, complaint, findings, work performed, deferred items, notes for next visit.
Page nesting goes as deep as you want. A common shape: Customers → Acme Construction → 2018 F-150 #4 with sub-pages for Service History, Inspection Photos, Customer Conversations. Or for individual customers: Customers → Smith, J. → 2022 RAV4. Both work. The point is that the page is a real document — you can paste in photos, lab results, scope captures, anything that needs to live alongside the written notes.
Plain markdown means the page is searchable, copyable, and exportable. A customer asks for their service records when they sell the car? You export the page as text and email it. Another shop asks for context on a referral? Same thing. There's no proprietary format to wrestle.
The agent reads across the whole shop history
Here's where the working memory of the shop becomes actually accessible. Capy, the assistant inside Docapybara, reads across your entire vault when you ask a question. So you can ask things like:
- "Have we ever seen this transmission code on a similar truck? Pull every '11–'14 Silverado with a P0700-family code." The agent searches every vehicle page, returns the matches with the relevant diagnostic notes from each.
- "What did the customer say about the brake pulsation last time they were in?" It finds the vehicle page, locates the customer-conversation note from that visit, and quotes it back.
- "List every warranty job we've done for this customer in the last two years and what the outcomes were." Cross-customer, cross-vehicle, in one query.
This isn't anything magical. It's the agent reading the same notes the techs wrote, doing the cross-reference faster than anyone could open twelve invoices in a row. The point isn't AI making decisions — it's AI removing the friction between "I think we've seen this before" and "here's exactly what we found and what fixed it."
Recording the customer write-up so detail isn't lost
The most diagnostically useful information in a service write-up is often the conversation, not the line items. The customer mentions the noise only happens when the AC is on. They say it started after the road trip. They mention they had it at another shop first who told them it was a wheel bearing.
Docapybara records audio inside the workspace and transcribes with speaker labels — so you can see what the service writer said and what the customer said separately. For complex driveability complaints, intermittent issues, or any vehicle coming back for the second or third visit, having the full conversation captured beats relying on a one-line write-up.
The transcript drops onto the vehicle page. When the tech opens the work order an hour later, they have the customer's actual words about the symptom — not a paraphrase that lost the most useful detail.
A live database of open work and pending parts
A :::database::: directive embeds a live database directly inside any markdown page. So your daily-board page can include a table of every active job with columns for vehicle, customer, work owed, parts status, ETA, technician, and notes. Six column types are available — enough for most shop tracking.
The table sits inside the page next to the written context for the day. Sort by ETA and you see the schedule pressure. Filter by parts status and you see what's on a truck and what's on backorder. When you tell the assistant "mark the rear pads on the F-150 as installed and add the alignment as a deferred item", it updates the row and makes the deferred-item note on the vehicle page in one step.
For parts management specifically, a similar inline database tracks pending orders — supplier, part number, ETA, vehicle it's for, status. Ask "what parts are we waiting on for jobs scheduled this week?" and the agent reads the database and tells you. Ask "which supplier has been slowest on transmission parts lately?" and it scans across the parts history and gives you the pattern. The agent-acts-on-docs side of this is the differentiator we describe in Claude Code for Documents.
Old service records and PDFs come along
Most shops have years of paper service records, scanned receipts, and shop manuals in PDF form sitting on a network drive that nobody opens. Drop those PDFs into Docapybara and the conversion pipeline turns each one into searchable markdown. The agent can now search across them the same way it reads your hand-written vehicle pages.
When you ask "what does the manual say about the transmission solenoid replacement procedure on this vehicle?", the agent pulls the relevant excerpt from the PDF as text. When you ask "have we serviced this vehicle in the last five years?", it scans every legacy invoice that was scanned in and gives you the history.
This matters most for older shops with deep archives. The institutional memory you've accumulated over the years stops being a closed paper file and becomes part of the searchable vault.
Operational notes — supplier accounts, equipment, training
A shop runs on more than vehicles. Supplier accounts, equipment service intervals, the lift inspection log, the new-hire training packet, the warranty-claim process, the after-hours emergency policy — all of that lives somewhere, and in most shops it's a folder of Word docs that nobody updates.
In Docapybara, those operational notes are nested markdown pages. A common layout:
Operations → Suppliers with one page per supplier (account number, terms, contact, return policy notes)Operations → Equipment → Lift Inspection Log as an inline database tracking date, inspector, condition, next-dueOperations → Training with pages for new-hire onboarding, specific procedure walkthroughs, certification-prep notesOperations → Warranty with the SOP and a database tracking active claims
When you onboard a new tech, you point them at one nested folder and they have everything — written process, live trackers, and the conventions the shop uses for vehicle notes. When you ask the assistant "what's the lift due for inspection next?", it reads the database row and tells you. When you ask "what did we learn from the last warranty claim with NAPA?", it finds the relevant note. For the SOP layer specifically, Standard Operating Procedures, Without the Wiki Maintenance Tax covers how to keep the written process from rotting.
Try Docapybara free
The fastest test is to run it on this week's schedule. Open Docapybara, create a page for tomorrow's first three vehicles, paste in whatever notes or photos you've got from their last visit (or just the work order), and ask the assistant for a one-paragraph summary of what to look for at each one. Five minutes of setup, and you'll know whether having the context next to you instead of buried in invoices changes the diagnostic process.
Try Docapybara free — bring a few of your messiest customer-history notes, last week's work orders, and one operational SOP you've been meaning to clean up. See how the workspace handles them.