Forge Care · Globe Routing Prepared for James Foster, Ph.D. · following your note of September 1
Globe Routing went from an idea in a Claude project to the tool your recovery team runs on, and it got there on your builds, our deployments, and a lot of evenings. You said it plainly: it needs to be supported in a more coherent and planned way, and it needs to be ready before several thousand pieces of archival tissue show up at the door. Here's how we'd do both, and exactly what it costs.
Every row after 0721 went through the same gate: diffed against last-known-good before it shipped. That gate is why production held through August. Forge Care makes it permanent.
The principle
"A working prototype and a production app are separated by the unglamorous parts: the diff, the staging pass, the permissions, the release note, and hours that belong to it instead of hours that get found. You've built something real. Forge is our name for carrying that last mile, every time, without taking the build away from you." Joshua R. Easters, Managing Partner, Diversicom
What you build
What we forge
In July a rebuild quietly dropped the pending-clear action, the Globe report, and the dashboard auto-expand. You noticed two. Now every bundle is compared against the last good build first, and dropped features are listed before anything ships.
The staging link you already use, loaded with your mock referrals and project scopes, so the live lists never take a hit while the kinks are worked out. Production moves when you say so.
When recovery staff are blocked, that's an outage, not a change request. We start within two hours of hearing about it, and you hear back the same day with what happened and what changed.
A build with a date on it gets its own room in the calendar: October and November at a higher hours level, the archive live by November 20, and December back to normal. No separate project fee.
From your note
First, credit where it's due. You built Globe Routing yourself, kept building it, and got your recovery team using it for real work. That's your app, and nothing on this page changes it. What your September 1 note changed is the pace: things have accelerated, and support that's been goodwill and evenings isn't a plan for a system your staff rely on.
Production
Lite and admin, live
On LWVI's own SharePoint and Azure, inside your tenant. Your code, your environment, your data
Staging
0831c, your test data
ParentProjects and the screening updates exercised without touching the live lists. Production held on 0818 until you say go
The rule
Diff before deploy
Since the 0721 rebuild dropped three features, every bundle is compared against last-known-good first
The history scan
Runs in the browser
Against a Keywords list you control. Nothing leaves LWVI, which is why the AI question can wait
The iPads
A friendlier address
A short name on the LWVI domain for the lite app, with the old link kept alive so nothing bookmarked breaks
This week
Real support traffic
The overnight push, a login error fixed the next morning, Shelley's loading issue. What an app in production generates
The work
Forge Care is a block of engineering hours that belongs to Globe Routing every month, with rules about how they're used. The archive intake runs inside it, at a higher level for two months, rather than as a separate project fee. No setup fee. No per-incident charges. Month to month.
$1,200at 10 hours a month
Hours scheduled in the workweek, not squeezed into the evening. Josh stays your named engineer, because he knows the code, the lists, and the history. A second engineer is briefed on the environment so a deployment never waits on one person's calendar.
Every hour is logged against the app and shown to you at a 15-minute monthly check-in: what shipped, what's queued, what the hours went to. Quiet month, hours don't pile up. Big month, we tell you before we go over, and you approve it first.
$1,80016 hours a month, two months
The archive intake has a date on it, so it gets its own room: the Build level for October and November, then December drops back to Active. Roughly thirty hours of build, inside the monthly, with a dry run before the real import.
If the build runs longer than we planned, that's ours, not a third month at the Build level. You see a one-page design before anything is built, and the archive is live and tested before the first box is opened.
Depends on one thing from your side: the manifest format from the collaborator, or a sample of it, by the design week. The earlier we see it, the less we guess.
On us, as things have been. This week's fixes, Shelley's issue, the 0831c staging pass, and the archive design week all happen before the first invoice.
Forge Care starts at the Build level. The archive list and the manifest import take shape on staging. Everyday support keeps running alongside.
Dry run, then production by the 20th. The guide updated, the recovery staff walked through it, the boxes can open.
Back to Active. If it's quiet, step down to Light at six or eight hours. If the next initiative shows up, step up. Thirty days' notice either way.
The release track
You send a bundle; this is what happens to it. On the left, how it has gone since June, honestly. On the right, how it goes under Forge Care. Nothing on the right is hypothetical. Most of it already happened for the 0831c bundle. The difference is that it's promised.
You build in your Claude project, export, and send it over with a note explaining what changed and, lately, test data to prove it.
Sometimes the same afternoon, sometimes after a sick week. You've written "not sure if you did any work on the files from last week" more than once. That's on us.
A bundle from you is acknowledged the day it arrives and on the staging link within two business days, with the diff results attached. If it needs to wait, you know why and until when.
What you'd notice A reply with a date in it, not a thread that goes quiet.
AI rebuilds branch from whatever base the model had in front of it, so a new bundle can quietly lose things an older one had.
The 0721 rebuild dropped the pending-clear action, the reformatted Globe report, and the dashboard auto-expand. Two of the three were noticed, one wasn't, and the graft took an extra round.
The new build is compared line by line against the last known-good, using the markers we recorded for the 0720 baseline. Anything missing is listed before it ships, and goes back into your source so it can't drop again.
What you'd notice A short list, or the words "nothing dropped," before you're asked to test.
The staging link loads your mock referrals and project scopes so the change can be exercised the way your team will use it.
When a change looked small, it sometimes went live directly. Small changes are where the login error and the lite-app loading issue came from.
Every build, however small, runs on staging with your test data. SharePoint list and column changes move with the code, so the app and its data are never out of step. Production waits for your word.
What you'd notice The same staging link, always current, and live lists that never get touched by accident.
The build goes live on LWVI's Azure static app and SharePoint, and your recovery staff pick it up the next morning.
"Thanks for burning the midnight oil" was kind of you. It also meant the first person to find the login error was you, at 9 AM.
Production pushes happen in the workweek, with a login and a recovery-queue check before anyone is told. If something changes for the staff, they hear it from the guide, not from a broken screen.
What you'd notice Fewer surprises before coffee.
Every production push gets a short note: what changed, what to check, what's next. Every hour gets logged against the app.
The record of what shipped lives in an email chain called "Globe Routing is live," which is now very long. Nobody could tell you how many hours it has taken. Including us.
The release note is written for Betty and Jason as much as for you. The monthly check-in shows the hours used against the block and what's queued, so the invoice never surprises anyone.
What you'd notice A paper trail that makes the app look as serious as it is.
Questions you might have
The things a scientist who built his own app would reasonably want to know before putting a support line under it. Filter by area.
Support
No. You keep building in your Claude project exactly as you do now. We take what you send and make sure it's safe to run, gets to staging and production cleanly, and stays secured. The code, the lists, and the IP stay LWVI's.
Your app, stillSupport
It's acknowledged Monday and on staging by Wednesday, with the diff. If your recovery staff can't use the app, that's different: same business day, and we start within two hours of hearing about it, weekend included.
Written targetsSupport
Time our engineers spend on Globe Routing: diffs, staging, deployments, SharePoint changes, fixes, access, the monthly check-in. Not the everyday IT back-and-forth, which stays part of being your IT partner, same as always.
Logged, shown monthlySupport
No. The app runs on your Azure and SharePoint. The history scan runs in the browser against your Keywords list. No AI service touches donor records, and that stays true until you ask for the compliant path, which is a separate quote you'd see first.
Inside your tenantThe archive
Most are. That's why the design week comes first and the import gets a dry run on staging. We map it once, load it as a job, and you check a sample before the real thing. If it takes longer than planned, that's ours.
Dry run firstThe archive
No. Search and query live inside the admin build, with the same look and logic as the rest of Globe Routing, and the archive rolls up through the ParentProjects list you already built. Your staff learn one tool, not two.
Same appThe archive
The list exists from the design week on, and staging can hold real records early if it comes to that. Worst case, the boxes wait a week in the freezer while the import finishes. Nothing is entered twice.
Honest answerMoney
July priced the compliant AI path: $5,000 to stand it up and $2,000 a month with the processing included. You deferred that, so it's gone from the number. What's left is the support, and it should cost less. It does.
AnsweredMoney
Step down to six or eight hours with thirty days' notice. Month to month from day one, no term, no minimum. If it ever stops earning its keep, you stop it, and the app keeps running where it is.
No lock-inThe Diversicom difference
Plenty of people can write code. These are the things that make Globe Routing's life easier day to day, and they come from twenty years in the building, not from a line item.
1410 N. 21st, Tampa. When the archive design needs a whiteboard, or the lab manager wants to see the list before it's real, it's a walk down the hall, not a Teams invite.
Every build since June has gone through our hands. The 0720 baseline markers, the shared-JS refactor, the ScannerFeedback list. Nobody re-learns your app on the first ticket.
The tenant, the SharePoint, the Azure subscription, the data warehouse the app sits next to. Globe Routing gets supported by the people who already secure everything around it.
No new vendor, no new account, no purchase order to chase. Forge Care shows up as a line on the Diversicom invoice LWVI already pays, and the hours behind it are on the record.
Where it runs
Nothing about Forge Care moves Globe Routing anywhere. It stays on the Azure static app and the ResearchSP SharePoint LWVI already pays for, the staging link stays where you bookmarked it, and the people supporting it sit one floor away. The only things that change are the calendar and the promises.
Proof, from inside LWVI
Forge Care isn't a new idea for us. It's the name for the way LWVI's custom tools have been carried for years, now with hours and promises attached.
Call Archive Search
A search layer over two decades of recorded calls, built and deployed in LWVI's own Azure with the files never leaving where they live. The same shape as Globe Routing: a real tool, on your tenant, supported by us.
The Webex cutover
Every phone number across ten states moved without a missed call, with a quick-start guide staff actually used. The same release discipline this page promises for your app, at organization scale.
June to September
Your app is the best reference on this page. Fourteen bundles, one regression caught, production stable through the busiest month, and a recovery team that uses it every day. Forge Care is that, with a calendar.
Investment
In July we quoted the version of this with the AI layer: $5,000 to stand up the compliant path and $2,000 a month with the processing included. This is the version without it, and it should cost less. It does. No setup fee, and even the two archive build months come in under that monthly.
Light
For a quiet stretch, once the archive is in
$800/month
Six hours a month. The floor, not a downgrade: the same engineer, the same release discipline
Active
The regular level for an app in real use
$1,200/month
Ten hours a month. Where December lands, and where Globe Routing has actually been running
Build months
October and November, for the archive
$1,800/month, two months
Sixteen hours a month covers the five-piece archive build. No separate project fee, and overrun is ours
Through December, by the numbers
≈ $4,800
October and November at the Build level, December at whatever regular level you pick on the right, Active unless you say otherwise. For comparison, the July proposal through December would have been $11,000 with the setup fee.
Nothing on this page invoices before October 1. September's fixes and the design week are on us, as things have been.
Beyond the block: $150 an hour, and only after you've approved it. Hours don't roll over, and we tell you before we'd go past them.
Same app, same people, a different scope. Every line on the right is on this page.
| What you get | July, with AI | Forge Care |
|---|---|---|
| Setup | $5,000 one-time | $0 |
| Monthly | $2,000, all in | $1,200 at the Active level |
| Included hours | Up to 10 | 6 to 12, your call each month |
| AI history-scan processing | Included, on our compliant account | Not needed yet. Separate quote when you're ready |
| Response | "Priority response" | Targets in writing: same business day when staff are blocked |
| The archive intake | Not in scope | Inside the monthly, two Build months at $1,800 |
| Release discipline | Implied | Diff, staging, production, release note. Every build |
| Term | Ongoing | Month to month from day one |
The July number isn't wrong; it priced a different thing. When the AI path is back on the table, that quote is the starting point, and Forge Care is what it would sit on top of.
Straight talk
Transparency up front. The numbers above rest on the assumptions below. We confirm each on the walkthrough and talk through any material change with you before proceeding.
Next steps
Thirty minutes, this week if you can. We walk through the plan and the numbers, you tell us what the archive really looks like when it lands at the bench, and we put the design week on the calendar. If something here is sized wrong, say so. The point of the hours block is that it fits the app, not the other way around.
Prepared September 2026 · pricing valid for 60 days from delivery.
