Claude Code set up for your repo, or an App Store preview video built from your screens. Flat price and a written scope. You know what you're getting before you pay.
People arrive needing one of three things: to read a real setup before paying anyone, one repo set up without doing it themselves, or the setup that outlives the session it was made in. Each paid rung is a flat price with a written scope and a clock. You can stop at any rung — none of them is a trial that expires.
Last updated: 2026-09-21
Not a developer? Start here instead: Claude Code for non-developers — what a CLAUDE.md is, why it keeps asking permission, and what a finished setup actually contains, in plain language.
Want to see the work before you pay anything? The starter repo, a real audit we published, a working MCP server with 15 tests that run without an API key, two files out of the Skills & Hooks Pack printed in full with its whole file list, and the free plugins are all public. Nothing is gated behind an email.
Before: you have seen CLAUDE.md mentioned and never seen one that a working repo actually runs on.
After: a real CLAUDE.md, a permission allowlist and three working skills, cloned and read in an afternoon.
Three free plugins install on their own too: kit-plugins.
Before: Claude Code asks permission before every command and re-learns your stack each session.
After: a CLAUDE.md written from your code, a tool allowlist and one skill, handed back as a PR you read and merge.
What lands, file by file: Setup Lite.
Before: one repo was not the problem — the setup dying with the session was.
After: a CLAUDE.md, a permission allowlist, hooks and 3–5 skills committed to your repo.
Shape of the output: kit-claude-code-starter. Need a tool built? That is MCP Basic $199, further down the ladder.
Shipping an App Store listing, not a repo? That is a different line of work and not a rung on this ladder: the ASO Launch Pack — $19 for the DIY templates, emailed the same day, or $149 done from your screens in 2 business days. Buying one does not start you on the Claude Code path above.
The three rungs above are where most people start. This is the full Claude Code path in order, so you can see what is above whatever you are looking at — and stop wherever it stops being worth it. Every row is a flat price with a written scope. None of them is a subscription and none of them expires.
| Rung | Who it is for | Price | |
|---|---|---|---|
| Starter repo | You want to read a real CLAUDE.md, allowlist and skill files before paying anyone anything. | Free | Clone it |
| Skills & Hooks Pack | You are happy to do the adapting. You want 10 skills and 8 hooks to drop into .claude/ today rather than write them over an afternoon. |
$29 | Buy the Pack $29 |
| Dual Setup: Cursor + Claude Code | You run Cursor and Claude Code on the same repo. You want one set of rules both of them actually load, and a check that fails the day they drift apart. | $29 | Buy the Dual Setup $29 |
| Setup Lite | One repo, and the thing you repeat every week. You want it done, but you do not want to spend $99 finding out whether this is worth it. | $29 | Buy Setup Lite $29 |
| Setup Sprint | One repo was not the problem — the setup dying with the session was. You want hooks, 3–5 skills and a runbook that outlives you. | $99 | Buy Setup Sprint $99 |
| MCP Basic | Configuration is not the gap. You need a tool built — one MCP tool with a schema, a smoke prompt and an off switch. | $199 | Buy MCP Basic $199 |
| DFY Async Repo Wire | You would rather not read a PR and decide. Hand over the repo, get it configured, plus .mcp.json stubs and a walkthrough video under 20 minutes. |
$249 | See the scope |
| Build Packet | An MCP server up to 5 tools, or a Worker up to 3 routes, with tests and a README. This is software, not setup. | $399 | Buy Build Packet $399 |
Why $249 starts with a form and not a checkout. Only two of those run at a time, so we confirm a slot before you pay rather than take the money and give you a date. Everything else on this list you can buy right now.
Already bought one and want the next rung? Reply to your receipt and say which one — we will tell you what it costs from where you are before you pay anything again.
The $19 ASO Launch Pack and the $149 preview packs are a different line of work — App Store listings and preview video, not Claude Code setup. They are not rungs on this ladder and buying one does not start you on it.
The difference is not effort and it is not quality. It is how many decisions are already made. Lite does the one thing you can name. Sprint is for when the honest answer to “which skill do you want?” is “I do not know yet, that is the problem.”
| Setup Lite — $29 | Setup Sprint — $99 | |
|---|---|---|
| What arrives | A CLAUDE.md, a permission allowlist, and one skill for the thing you repeat most — as a pull request you read and merge | The same, plus 3–5 skills, hooks, up to 3 MCP servers wired, and a one-page runbook |
| How many repos | One | One, gone through properly rather than more of them |
| Clock | 24 hours from a complete intake | 48 hours from a complete intake |
| Take this if | You can finish the sentence “it should stop asking me about ___ and it should know how to ___” | You have tried to set it up, it half works, and the files now disagree with each other |
| Not for you if… | …you want hooks or MCP servers. Lite has neither, on purpose, and buying it twice is worse value than Sprint once | …you have not run Claude Code on this repo yet. Spend the $29 first, or nothing, and come back knowing which five skills you actually wanted |
The one-line version: if you know what you want done, take the $29. If what you want is for the whole thing to stop fighting you, take the $99. And if it is already set up and still fighting you, neither of these is the purchase — that is a Surface Audit $79, which reads what you have and hands back the contradictions.
Starting at $29 is not a wasted step. Everything in Lite is in Sprint, so the work carries; the ladder is built to be climbed one rung at a time rather than guessed at from the bottom.
Buy Setup Lite ($29) if you can finish the sentence “it should stop asking me about ___ and it should know how to ___”. You get a CLAUDE.md, a permission allowlist and one skill, back as a pull request in 24 hours. Buy Setup Sprint ($99) if you have tried to set it up, it half works, and the files now disagree with each other — that adds hooks, 3–5 skills, up to 3 MCP servers wired and a runbook, in 48 hours. If it is already set up and still fighting you, neither is the right purchase: that is the Surface Audit $79, which reads what you have and returns the contradictions and a fix list. Starting at $29 is not a wasted step, because everything in Lite is also in Sprint.
Plenty of people buying this are not senior engineers. Product people, designers, ops, solo founders shipping something they mostly vibe-coded. If that is you, the rungs above start one step too late. Here is the earlier path.
First, for nothing. Install the starter setup and read it. It is a real CLAUDE.md, a real allowlist and three real skills, not a brochure: kit-claude-code-starter. Three plugins are free too and each installs on its own — ship-check (catches the things that get an app rejected), explain-diff (says in plain words what a change actually did), and storekit-check. If that is all you ever take from here, that is a fine outcome.
Then, if you would rather not do it yourself. Setup Lite, $29, is the small version: one repo, a CLAUDE.md written for what is actually in it, a permission allowlist so it stops asking before every command, and one skill for the thing you do most. It comes back as a pull request you read and merge. No call, and you do not have to know what any of those words mean before you buy — the intake form asks in plain language.
When one repo and one skill is not enough, Setup Sprint, $99, is the same idea with room in it: 3–5 skills, hooks, up to 3 MCP servers wired, and a runbook so the setup outlives the session it was made in.
Already set up, and fighting it? That is not a bigger purchase, it is a different one. Surface Audit, $79, reads the CLAUDE.md and settings you already have and hands back the contradictions and an exact fix list. Here is a real one, published in full — an audit of our own starter repo, findings and all.
What is deliberately not on this path: the MCP and Build tiers further down the page. They are good, and they are not where anyone should start.
The Same day rows are downloads: the clock starts at your Stripe receipt and there is no form to fill in. Every other clock starts at a complete intake, the form further down, not the receipt. If something is missing we ask once, and the clock starts when you answer.
| What you buy | What comes back | Clock |
|---|---|---|
| ASO Launch Pack: DIY — $19 | Playbook and paste-ready templates, as a download | Same day |
| MCP Config Pack — $49 | 8 forkable mcp.json configs, a smoke prompt each, install README | Same day |
| Dual Setup: Cursor + Claude Code — $29 | Shared CLAUDE.md + Cursor rules block and a drift check, as a download | Same day |
| Skills & Hooks Pack — $29 | 10 skills, 8 hooks, a wired settings.json and a hook test script, as a download | Same day |
| Permit Ready Pack — $39 | Call script, document checklist, inspection walks and corrections log, as a download | Same day |
| Sub Intake Vault — $49 | Sub evidence register, change order log, daily field log, as a download | Same day |
| Setup Lite — $29 | CLAUDE.md, tool allowlist and one skill, handed back as a PR | 24 hours |
| Claude Code Surface Audit — $79 | Written audit, contradiction map, exact fix list | 48 hours |
| AEO Visibility One-Shot — $99 | 10 queries checked in AI answers, 5 pages audited, two rewrites, fix list | 2 business days |
| Setup Sprint — $99 | CLAUDE.md, allowlist, hooks, 3–5 skills, up to 3 MCP servers, runbook | 48 hours |
| MCP Basic — $199 | One MCP tool, schema, smoke prompt, handoff notes | 72 hours |
| DFY Async Repo Wire — $249 | Tuned CLAUDE.md, .mcp.json stubs, permissions note, 20-min walkthrough | 5 business days |
| ASO Launch Pack: Done for you — $149 | Listing copy and stills made from your screens | 2 business days |
| Preview Pack — $149 | Preview video to Apple’s spec, 5 stills, 2 revision rounds | 3 business days |
| Preview Pack+ — $249 | Preview Pack plus 3 caption variants, 2 revision rounds | 3 business days |
| Build Packet — $399 | One MCP server (up to 5 tools) or one Worker (up to 3 routes), tests and README | 5 business days |
One project at a time, so start dates are first come, first served. A scope change is a new order, not a revision round.
Try it first: free plugins, including a StoreKit pre-flight check, at github.com/sdvsignal/kit-plugins.
See a sample: an iOS app backend Worker you can deploy in one click.
See a sample setup first: kit-claude-code-starter. Or read a sample config audit of it. One project at a time, so start dates are first come, first served.
Before: Claude Code asks before every command, forgets your stack each session, and the one CLAUDE.md you wrote in a hurry is not enforced by anything.
After: a configured repo — handed over on a branch, or as a zip if you would rather not give access — plus a one-page runbook and a file listing what was built against what you ordered.
| What you get | What is in it | How it was checked before you saw it |
|---|---|---|
CLAUDE.md | Your install, dev, test and lint commands, your “never do this” list, and the traps a new contributor hits in this repo | Every command was run in your repo and passed |
| 3–5 skills | Drawn from the work you actually repeat, not from a list of nice ideas | Each one invoked once to prove it does what it says |
.claude/settings.json | A permission allowlist so normal work stops prompting, and hooks — at minimum your formatter running after every Edit and Write | A normal session run end to end to confirm the prompts actually stopped |
.mcp.json | Up to 3 MCP servers wired, credentials referenced from env vars and never embedded | Each server connects, and how to remove it is written down |
| 1-page runbook | Gets a new dev going in about 15 minutes | — |
ACCEPTANCE.md | What was built, line by line, against what you ordered | This is what the revision round is measured against |
Then a doctor.sh pass runs against the whole thing and everything it flags is fixed before you see it. Nothing in the delivery carries a secret, a token, or an absolute path from our machine. Those checks are the $99.
Your part: a repo URL with collaborator access to a branch, or a zip. Your stack and the four commands if you know them, your never-do list, which MCP servers you want if you know, and how you run Claude Code. “Not sure” is an answer to any of those. No call, no screen share. We never ask for an API key or a .env; if you send one by accident we will tell you to rotate it and keep it out of the work. 48 hours from a complete intake, not from the receipt. One paid job runs at a time, so yours is the only repo open while it is being built.
What it is not: not work on your default branch, not a custom MCP tool — that is MCP Basic $199 or Build Packet $399 — and not more than the list above; anything past it is a new order. If all you need is your commands and one skill in a pull request, Setup Lite $29 is the right buy and we will say so rather than stretch this one.
Lite gets your commands and one skill into a pull request and stops there. Sprint is the part Lite tells you it is not: the allowlist so prompts stop, hooks, 3–5 skills, up to 3 MCP servers, the runbook and ACCEPTANCE.md. If what bothers you is only that Claude does not know your repo, buy Lite. If what bothers you is that it asks before every command and nothing runs after it edits, that is this.
Because they do not know your repo. kit-claude-code-starter is a complete, generic setup and kit-plugins has free plugins, including a StoreKit pre-flight check. You fill in your commands, your traps, your servers, and you run the checks yourself. If you would rather do that, take them. They are free and we would rather you had them than nothing.
Leave it blank or say so. We recommend from your stack instead, wire what makes sense within the cap of three, and the removal path for each one is written down — so a wrong guess costs you a file delete, not a rebuild.
Write “not sure” and send the repo. Working the commands out of your code is the job at this tier, not a gap in your intake. You will not get an email back demanding the four commands before anything starts.
Collaborator access to a branch, and nothing else — or send a zip and get a zip back with the same files. MCP credentials stay in env vars on your side, referenced from the config and never written into it.
Tell us and we fix it against the written scope — one round, included. ACCEPTANCE.md lists what was built against what you ordered, so the conversation is about a list rather than a feeling. Things outside that list are a new order, not a revision.
Both, and we would rather say so. AI tools including Claude produce the work, and a person reads every deliverable and runs the checks: the commands in your repo, a session to see the prompts stop, each skill once, each server connecting. You are paying for the checks more than the typing.
No. Everything is plain files in your repo. The runbook is one page, each MCP server has its removal path written down, and nothing points back at us.
Sprint configures Claude Code around a repo you already have; it does not build anything new. If you want a custom MCP tool, that is MCP Basic $199. Bigger than that, Build Packet $399. If you order Sprint and describe a build, we will say so before starting rather than stretch the scope.
Before: the thing you need does not exist yet, and the quotes you have back are "let's scope it" and a call.
After: one built, tested, deployable thing, handed over with the tests and a README, in five business days from a complete intake.
It is one or the other, not both. Pick the shape at intake:
| Shape | Ceiling | What comes with it |
|---|---|---|
| One MCP server | Up to 5 tools | Described schemas, the unhappy paths handled per tool, a smoke prompt each, install and removal notes |
| One Cloudflare Worker | Up to 3 routes | Tests, deploy config, error shape with a request id, and the removal path |
Five business days from a complete intake, not from payment. One round of fixes against the written scope. One project at a time, so start dates are first come, first served.
The Worker shape: kit-ios-worker-template — a StoreKit 2 subscription backend with a Deploy to Cloudflare button, D1 migrations, App Store Server Notifications V2 handled idempotently and out-of-order safe, a per-device KV rate limit returning 429 with Retry-After, and every error as JSON carrying a request_id. npm test runs 5 unit tests offline; there is a Playwright end-to-end suite in the repo as well.
The MCP shape: kit-one-tool-mcp — 15 offline tests, mostly failure paths, with three standing up a real MCP client against the real server. That repo is one tool; a Build Packet is up to five of them built against your API.
Read the samples honestly: the Worker template has five routes, which is more than the three a Build Packet buys. It is published as a quality bar and a starting shape, not as a scope statement — it exists so you can judge the tests, the error handling and the README before you spend $399, not to imply you get that exact thing. What you get is built against your API, at the ceilings in the table above.
Your part at intake: which shape, what it should do, the API docs, the auth type, and a sample request and response or a sandbox. Credentials stay yours — env vars on your side, a revocable sandbox key if we must test, never a live secret in a form. Delivered as a private repo invite or a zip from support@sdvsignal.com.
What it is not: not a retainer and not ongoing maintenance — there is nothing to cancel. Not hosting: a Worker deploys to your Cloudflare account and the bill is yours. Not a rewrite of an existing system. Not an app, a front end or a design. If one tool is all you need, MCP Basic $199 is the honest purchase and we will say so. If you want your existing repo configured rather than something new built, that is Setup Sprint $99. If you want a repo wired for Claude asynchronously, that is DFY Async Repo Wire $249.
Because that is two builds and it would not land in five business days. The ceiling exists so the date means something. If you genuinely need both, buy one now and the other after — or say so at intake and we will quote it properly rather than guess a price and miss.
Nothing is skipped; the scope is narrow, which is the whole point. There is no discovery phase, no call, no design round and no stakeholder review, because the shape and the ceiling were agreed in writing at intake. The clock starts when the intake is complete, so a missing API doc pauses it rather than compressing the build.
You do, and it is your account. A Worker is deployed to your Cloudflare, with your D1 and your KV, so you keep the keys and the bill and you can revoke our access the day it is handed over. We do not host anything for you and there is no subscription attached to this.
You have the tests and the README, and it is plain code in your repo. The one round of fixes covers anything that does not match the written scope. Beyond that it is a new order — there is no support plan to buy, and we would rather say that plainly than sell you a retainer you do not need.
This rung builds one new thing and hands it over; it is not a refactor of what you already have. A Worker can sit in front of an existing API and an MCP server can call one. If the job is really "fix and extend what exists", say so at intake — that is not this, and we will tell you before you pay rather than discover it on day three.
No. Say what you want Claude to be able to do and we will tell you whether it is an MCP server, a Worker, or neither. If what you describe is smaller than this rung, you will be pointed down the ladder — MCP Basic $199 or Setup Sprint $99 — not sold the larger one.
| You send | You get |
|---|---|
| Screen recordings or screenshots, any size A two-line hook for each screen | Preview video (886×1920) 5 stills Captions with no prices and no Apple badge art |
Different from the Preview Pack: no video, just listing copy and stills. We never submit to Apple for you. DIY is a download, and we email your link the same day you pay.
Same pack, same files, same checklist. The only question is who fills the templates in. If you already know what your app is and who it is for, the $19 gives you the structure and you are done in an afternoon. The $149 exists for the case where the blocker is not the format, it is not knowing what to say about your own app.
| DIY — $19 | Done for you — $149 | |
|---|---|---|
| What arrives | The playbook and the blank templates — keyword map, subtitle, promotional text, description, screenshot storyboard, paste checklist. Yours to fill in | The same documents, already written from your screens and your category, plus 5 stills at 1290×2796 made from your captures |
| Who writes the copy | You do. The templates are blank until you write in them — that is the whole $130 difference | We do, and you get the reasoning next to each field so you can maintain it yourself next release |
| What you supply | Nothing. It is a download | Your app name and category, 5 to 8 portrait screen captures, who it is for in your own words, and your current listing plus a screenshot of your App Analytics baseline — that number is not recoverable later and it is the only way you will know whether this worked |
| When | Same day, usually within the hour. Bookmark the download link and fetch it again whenever you need to — links do get rotated now and then, and one reply to your receipt gets you a fresh one the same day | 2 business days from a complete intake — the clock starts when the captures land, not when you pay |
| Who submits to Apple | You do, both ways. We hand you files and a paste order; we never touch your App Store Connect account and we never submit on your behalf | |
| Not for you if… | …you were hoping to hand this off. Nothing is written for you at this tier — if you were not going to sit down and do it, the $19 becomes a zip you never opened | …you can already write your own store copy. Then you are paying $130 for typing you were capable of doing, and the $19 is the honest purchase. Also not for you if you wanted a video — that is the Preview Pack, which is confusingly also $149 and is a completely different deliverable |
The one-line version: if you know what to say and just want the structure, take the $19. If you know your app is good but go blank writing the subtitle, take the $149. Neither one is a ranking — nobody can sell you that, and anyone who says otherwise is guessing at an algorithm Apple does not publish. What you are buying is a listing that says the right thing in the right field, which is the part you control.
These are not Claude Code products and they are not software. They are plain
.xlsx workbooks and a PDF, for general contractors and remodelers who are losing time
to paperwork. Nothing here is a rung on the ladder above, and buying one does not start you on it.
They are on this page because the same person builds both.
The failure it is for: a sub's general liability lapsed in March, you found out in June, and the only reason you found out is that something went wrong.
The failure it is for: two weeks spent collecting documents nobody asked for, or reaching intake and hearing about a review you had never heard of.
Both are instant downloads — the link is on the page you land on after
paying, and the page stays bookmarkable. Plain .xlsx: opens in Excel, Numbers and
Google Sheets. No app, no login, no subscription, nothing to cancel.
Read this before you buy either one. These are operations templates. They are not legal documents and not legal advice, and they are not AIA documents or a replacement for one. The Sub Evidence Register is a tracker: it records what you typed. It does not verify a licence and it does not check a certificate of insurance against the carrier — a green row means the dates you entered are not in the past, and nothing more. Confirm licences with your state board and coverage with the carrier or your broker. The Permit Ready Pack never tells you what your jurisdiction requires; the whole design is to get you asking your AHJ instead of guessing. Requirements vary by state and by project and they change. No licensing board, insurer or department has approved, endorsed or certified any of this.
See what you get first: kit-one-tool-mcp is a working sample with 15 tests you can run without an API key.
Before: Claude cannot reach the one system your work actually lives in, so you paste things into the chat by hand and it never sees anything current.
After: one tool it can call, with a described schema, that fails in sentences instead of stack traces.
| You get | What that means in practice |
|---|---|
| One MCP tool against your API | Named from the caller's side, so Claude picks it for the right question |
| A described input schema | Every field documented, required vs optional explicit. A bad argument is rejected by the schema, not by your handler at runtime |
| The unhappy paths, handled | Bad auth, 429, empty result, network failure, non-JSON body. An empty result reads as empty, never as a crash |
| A smoke prompt | The literal sentence you type in Claude to make it fire, and what you should see back |
| An enable and disable path | The exact config to add, and the exact steps to remove it cleanly. The README tells you how to take it out, not only how to put it in |
| 72 hours from a complete intake, one revision round | The clock starts at intake, not at payment. 72 rather than 48 because the smoke prompt has to have actually fired against a real API |
One tool. Not one server with five tools. If what you describe needs a second tool, we say so at intake and point you at Build Packet $399 — before building, not at delivery. Upselling at delivery is a bait and switch and we would rather lose the order.
kit-one-tool-mcp is a real sample of what this hands over. git clone, npm install, npm test — 15 tests, all offline, no API key and no network. Most are deliberately the unhappy paths: missing key, 401, 403, 429, 503, DNS failure, non-JSON body, empty result. Three of them stand up a real MCP client against the real server over an in-memory transport and call the tool, so the wiring is tested rather than just the function.
That is the whole proof and there is no other. We have no case studies and no client logos. Read the repo, run the suite, and judge the shape of the handover — including the README, which is the part most samples skip.
Your part at intake: one sentence on what the tool should do, a link to the API docs, the auth type, and a sample request and response or a sandbox. We never take your credentials. The key lives in your env var, we never hold it, and the delivered repo does not contain it. If we need to test against your API, ask for a sandbox key you can revoke, and revoke it afterwards. Delivered as a private repo invite or a zip from support@sdvsignal.com.
What it is not: not a pack of configs for servers that already exist — that is MCP Config Pack $49. Not wiring existing servers into your repo — that is Setup Sprint $99. Not five tools and not a Worker — that is Build Packet $399. Not hosting, not monitoring, and not an ongoing integration.
Because "one tool" is a scope you can hold us to and we can finish. Most of the cost is not the function — it is the schema, the five failure modes, and actually firing it in Claude against a real API. A second tool doubles the surface, so it is priced as Build Packet $399 rather than quietly absorbed.
Because a passing unit test is not the bar here. The smoke prompt has to have been actually run in Claude Code or Claude Desktop with the tool firing against the real API or your sandbox, before delivery. That requirement is why the clock is 72 hours and not 48.
No, and please do not send one. Credentials are read from process.env on your machine — never a literal, never a default that happens to work. If testing needs real access, send a sandbox key you can revoke and revoke it when we are done. If you paste a live secret to us by accident we stop, tell you which field carried it, and ask you to rotate it.
Send the docs link at intake and we will say yes or no before you pay. No docs, no build — that is a firm rule, because a tool written against a guess cannot be tested. OAuth flows that need a browser round trip are usually more than this rung; we will tell you that rather than take the order.
It is yours. Plain files, your env var, and a README with the config to paste, the sentence to type to test it, and how to remove it cleanly. Nothing points back at us and there is nothing to cancel.
Anything that does not match the written scope agreed at intake. A field you did not mention, a second endpoint, or "can it also…" is a new order, priced, not a revision. We write the scope down at intake precisely so this conversation is short.
Both. AI tools including Claude write the server, and a person runs the checks — the suite, the smoke prompt in a real client, the bad-auth path, and a read of the README for anything that would leave you stuck. The checks are what you are paying for more than the typing.
Configs only, for Claude Code — we don’t install them on your machine. We email your download link the same day you pay. Bookmark it and re-download whenever you need to — links do get rotated now and then, and one reply to your receipt gets you a fresh one the same day.
A pack of “8 configs” is not something you can judge. Here is the actual list, so you can decide you already have six of them and not buy this.
| Server | Type | Why you would want it |
|---|---|---|
| filesystem | local | scoped file access outside the working tree |
| context7 | remote | current library docs instead of stale training data |
| playwright | local | drive a real browser for UI checks |
| XcodeBuildMCP | local | build, test and run iOS apps from Claude Code |
| github | remote (official) | issues, PRs, repos |
| sentry | remote (official) | errors and issues from production |
| supabase (read-only) | remote (official) | query your schema without write risk |
| stripe (test mode) | local | products, prices and test payments |
Every one was checked against that server’s own docs or npm package on 2026-09-19, and the two deprecated ones people still copy from old blog posts — the old @modelcontextprotocol/server-github and server-postgres — are deliberately not in it. Each config ships with a smoke prompt, so you can tell in one line whether it actually connected.
Honest limits, the same ones printed inside the pack: servers change and the pack pins nothing, so if one breaks, check its project page first. Remote servers log you in with the vendor’s own OAuth, and you paste your own tokens into your shell profile — no secrets ship in the files and Kit never sees yours. This is configuration, not a custom server: for one built against your API, that is MCP Basic $199 or Build Packet $399.
Already have an .mcp.json that works? Then skip this. It is eight files and a README, and the $49 buys the checking rather than the typing. If what you want is the wiring done in your repo, Setup Sprint $99 includes up to 3 MCP servers wired.
Both say “MCP” and they are not the same purchase. The question is not how much you want to spend, it is whether the server you need already exists. If it does, nobody should be charging you $199 to write one.
| Config Pack — $49 | MCP Basic — $199 | |
|---|---|---|
| What arrives | 8 mcp.json configs for servers that already exist, a smoke prompt for each, an install README | One MCP tool built against your API, with its schema, handoff notes and an enable/disable path |
| Whose service it talks to | Other people’s: filesystem, context7, playwright, XcodeBuildMCP, github, sentry, supabase, stripe | Yours — the one nobody has written a server for |
| Who writes code | Nobody. These are configuration files; the servers are already published | We do, and you get the source |
| What you supply | Your own tokens, pasted into your shell profile. No secrets ship in the files | Your API, plus an endpoint or docs good enough to build against |
| When | Same day, as a download link you bookmark and re-use — if yours is ever rotated, one reply to your receipt gets a fresh one | 72 hours from a complete intake |
| Not for you if… | …you already have an .mcp.json that works, or you only wanted one of the eight — several are free to set up in ten minutes from the vendor’s own README | …a public server already covers it. Check the eight above and the vendor’s docs first; if one of them does the job, buying this is $150 you did not need to spend |
The one-line version: if you can name the tool you want and someone else already publishes it, take the $49. If the thing you want Claude Code to reach is yours, take the $199. If you know which servers you want but not how to wire them into your repo, neither of these is it — that is Setup Sprint $99, which includes up to 3 wired.
Before: rules in your CLAUDE.md that your settings do not enforce, denies that miss the second spelling of the same command, and no easy way to tell which is which from inside the repo.
After: a written report, by email, where every finding names the file and the line or the settings key.
| You paste | Why it is needed |
|---|---|
Every CLAUDE.md — repo root, subfolders, ~/.claude/CLAUDE.md | A rule only holds if all of them agree |
.claude/settings.json, .claude/settings.local.json, ~/.claude/settings.json — tokens and env values deleted first | This is where a rule is actually enforced, or is not |
| Hooks — the commands, plus short script bodies | The third place a rule can live, and the one people forget to check against the other two |
.mcp.json with secrets as <redacted> | What your settings and skills are allowed to reach |
| One sentence on what annoys you most about Claude in this repo | The one thing the report is not allowed to miss |
The report has five fixed sections: Summary, Findings, Contradiction map, What’s already good, Optional next step. The contradiction map is the product — pairs of rules that cannot both hold, which is the thing you cannot easily see from inside your own setup. “Your CLAUDE.md is too long” is not a finding and does not get written. “What’s already good” is not padding; it is there so you do not delete something that works. Nothing is invented to fill a section.
The only proof on offer is public: a real audit of our own free starter, live at /samples/cc-surface-audit. Four files went in and five findings came out, two of them High. The headline one: the CLAUDE.md says “never force-push” while the settings deny only git push --force, so git push -f goes straight through. A report you pay for is not thinner than that one. That is the $79.
Your part: paste the five things above into the form below. No call, no install, no repo access. Delivered as inline markdown by email from support@sdvsignal.com within 48 hours of a clean intake.
What it is not: not repo access at any price at this rung, not a security certification, and not the fixes themselves — applying them is Setup Lite $29 or Setup Sprint $99. If the report names an MCP server you do not have yet, building one is MCP Basic $199.
Because at $79 the paste is the product. It is a scope ceiling and a liability ceiling at the same time: we only ever see what you chose to send. If you want someone working in the repo, that is Setup Lite or Setup Sprint, not this.
Five things. Every CLAUDE.md you have, including the one in your home folder. Your settings files, with tokens and env values deleted before you paste. Your hooks, commands plus the short script bodies. Your .mcp.json with secrets replaced by <redacted>. And one sentence on what annoys you most. If you do not have one of them, say so and skip it.
We stop. It is not saved, you get an email naming which field carried it, and we ask for a redacted resend. Rotate the key regardless. The 48-hour clock restarts from the clean copy, not the first one.
Not from a paste, no. The fix list is written so you can apply it yourself in one sitting. If you would rather someone did it in the repo, Setup Lite $29 covers a CLAUDE.md and one skill as a pull request, and Setup Sprint $99 covers settings, hooks and MCP as well.
Read the sample before you decide; it is the whole proof and there is no other. One finding there: rm -rf is denied and rm -fr, rm -Rf and rm -r -f are not. Another: a ship-check skill runs a type check, the settings allow it, and the CLAUDE.md never names a typecheck command — so the skill reports FAIL on every run and people learn to skip the FAIL rows. We have no case studies and no “clients often” line, and we would rather you judged the sample than took our word.
The starter we audited is small too: one CLAUDE.md, one settings file, one hook, one example MCP config. It still had two High findings. If you have a CLAUDE.md and a settings file, there is a contradiction map to draw. If you have neither yet, do not buy this — take the free starter and come back once you have changed it.
No. It is a configuration review. It will tell you when a deny rule misses a spelling, because that is a rule not doing what your CLAUDE.md says it does. It does not certify anything, and you should not show it to anyone as if it did.
Both. AI tools including Claude produce the draft, and a person reads every report before it goes out and holds it to the bar: every finding cites a file and a line or a settings key, and nothing is added to fill a section. A finding that cannot cite its line does not ship.
Because that is what a real audit of a real setup looks like. The starter is free and shipped with those gaps; the sample is the audit that found them. A sample with nothing in it would be a reason not to trust us.
Read the sample before you buy. A real one-shot we published is up at /samples/aeo-oneshot — same format you get back, nothing gated. Two business days from a complete intake.
Before: people ask an assistant the question your page answers, and the answer names someone else. You cannot see it happening and nothing in your analytics reports it.
After: one document that names the queries, names who got cited on each, and gives you the replacement copy — not advice about the copy.
| In scope | Why it is the part that matters |
|---|---|
| 10 queries, checked by hand in Google AI Overviews, ChatGPT and Perplexity, each with the source that got cited | This is the half a competitor cannot run for you from a script. It is also the page people forward to their boss |
| Up to 5 pages audited for extractability, schema, crawler access and machine-readable files | The mechanical half. Reproducible, and the checker is in your deliverable so you can re-run it |
| Two actual rewrites — your current paragraph, and the replacement | “Make this more extractable” is worth nothing. A 40–60 word replacement is worth the fee on its own |
| Ordered fix list with time estimates, and a list of what to skip | Ordered by effect per hour, not by where it appeared. The skip list is the part that makes the rest trustworthy |
| Delivered as markdown and PDF | The markdown is what you paste to whoever does the work |
No score. There is no “visibility score out of 100” in the deliverable. It would be unfalsifiable, and it is the first thing a sharp reader attacks. Findings cite a page and a check, or they do not ship.
We ran this audit on our own site and fixed what it found the same evening, then published the whole thing: /samples/aeo-oneshot. The scoreboard there is a real before and after across three pages — 7 FAIL findings to 0. Our two comparison pages carried no schema at all and were 181 and 178 words, too thin for anything to quote; they are now ~715 each, and the added words are a table and five real questions rather than padding. Question-shaped H2s went from 0 of 2 to 3 of 4. /pricing.md and /pricing.txt were both 404 to an agent asking what we cost; both now return 200.
Read the honest limit on that page before you judge it: the sample shows the mechanical half. The ten hand-checked queries — the judgement half, and the reason this is $99 rather than a script — are not in it, because publishing our own query results would tell you more about our competitors than about the product. The sample also has a “what we did not do, on purpose” section. That is deliberate: a fix list with nothing excluded from it has not been thought about.
Your part: a short intake after you pay — your pages, your buyers, and the questions you think they ask. We send the ten drafted queries back for a yes before auditing them, so two minutes of yours stops us auditing the wrong intent. The clock starts when that comes back. Delivered by email from support@sdvsignal.com.
What it is not: not implementation — it tells you what to change and hands you the copy, it does not push it to your repo. Not monitoring, and not a subscription; if you want it again in two months you buy another one. Not SEO in the ranking sense, and not a guarantee that any assistant will cite you. If what you actually need is the App Store rather than AI answers, that is ASO Launch Pack $19 / $149. If you need the fixes applied in a codebase, that is Setup Lite $29 or Setup Sprint $99.
You land on a thanks page and the intake form. Send the 10 queries you care about and the pages you want looked at; the two business days start when that intake is complete, not when the card clears. No call and no retainer — one report, then we are done.
It overlaps and we would rather say so. Schema, crawler access and clean structure help both. What is different is the query half: classic SEO asks where you rank on a results page, and this asks whether an assistant answering the question names you or names someone else. A page can rank fine and still never be cited. The fix list is ordered for the second problem.
Because they are checked by hand in three engines and recorded with the source that got cited, which is roughly half the fulfilment time. A hundred machine-pulled queries would be cheaper to produce and worth less: AI answers vary by phrasing and session, so the value is in choosing the ten that match real buying intent and reading what came back. If ten is genuinely too few for you, this is the wrong product.
Every row names the query, the engine and the source that was cited. A query we could not check is marked not-run with the reason, never filled in. Inventing a citation row is the one thing that would end this product, so the rule is written into the runbook rather than left to judgement.
No one can promise that and we are not going to. What the audit gives you is the specific reasons you are not being cited today and the copy that removes them. Our own before-and-after is on the sample page and it is a fix log, not an outcome claim — we have not published a citation-rate number because we do not have one worth believing yet.
Both, same as everything here. The mechanical checks are a script you get a copy of. The query work is a person in three engines, and a person writes the two rewrites and decides the order of the fix list. The parts that are judgement are the parts you are paying for.
Then read the sample first. Ours is a small site too — a home page and two comparison pages — and it had seven failures. But if you have one page and no comparison content, there is not much for an audit to work with, and you would get more from writing the second page than from buying this.
.claude/, plus a settings.json that wires them togetherThree of the eight hooks, refusing three real mistakes, then the pack’s own test suite (134 checks in the hooks suite; 176 across both). Every line is captured output, not a mockup — the block messages are the hooks’ real wording.
Two of the files are published in full before you pay: /samples/skills-hooks prints them end to end and lists everything else in the zip. If the two you can read for free are not worth $29 to you, the other 23 will not be either — that is the point of showing them.
This and Setup Lite are both $29 and they are not the same purchase. This is files you drop into your repo and adapt yourself: generic on purpose, nothing here has seen your code. Setup Lite $29 is a person writing a CLAUDE.md from your code, running every command in your repo first, handed back as a PR. Buy this one if you want to move today; buy that one if you would rather not do the adapting.
What it is not: not a framework and not a dependency — the files land in .claude/ and do nothing until you use them. No subscription, no licence server, nothing phones home. The hooks are shell scripts you can read in a sitting, which is the only reason to trust a hook that can block your commits.
Skills: onboarding a repo, test-first, a debug loop, safe refactors, dependency checks, a security sweep, hunting flaky tests, reviewing your own diff, PR descriptions and release notes. Hooks: block secrets, refuse reads of secret files, refuse destructive shell, protect a branch, warn on a huge diff, stop test-skipping, format what changed, and a session summary. The full file list is on the sample page.
The hooks are POSIX shell and assume git. The skills are markdown instructions, so they are stack-agnostic by nature — but a few name commands like npm test, and you change those to yours. INSTALL.md says which lines to edit. If you want that editing done for you against your actual repo, that is Setup Lite $29.
You could, and the sample page is most of the shape, so we would rather say that than pretend otherwise. What you are buying is the afternoon, and the eight hooks already having been argued with — a hook that blocks your own commits is easy to write and annoying to get right.
It is a download, so ask and you get it — email support@sdvsignal.com with your Stripe receipt. We would rather refund you than have you feel talked into a zip.
Cursor reads .cursor/rules/*.mdc. Claude Code reads CLAUDE.md. Nothing keeps them in agreement, so a month in, one of them is telling the truth about your build and you don’t know which.
Stop hook.md in .cursor/rules is ignored, and so is an .mdc with no frontmatterThree things here cost $29 and they are different purchases. This is a download of templates you fill in. The Skills & Hooks Pack is a download of a .claude/ folder, Claude Code only. Setup Lite is a service: we write your CLAUDE.md from your own code and hand it back as a pull request.
If your instructions are simple, Cursor also reads a plain AGENTS.md, and you can keep one file and skip this. We would rather say that than sell you something you do not need.
Before: Claude Code asks permission before every command and re-learns your stack each session.
After: three files, in a pull request you read and merge.
| File | What is in it |
|---|---|
CLAUDE.md | Your install, dev, test and lint commands, your “never do this” list, and the two or three things a new contributor gets wrong in this repo |
.claude/settings.json | A permission allowlist for the read-only commands your repo actually runs, so the prompting mostly stops |
| One skill file | The thing you repeat every week, written so you can run it by name |
Every command written into that CLAUDE.md was run in your repo and passed before the PR opened, and the skill was invoked once to prove it does what it says. That check is the $29.
Your part: three fields on the form below — where the code lives, how it gets run, anything to keep hands off — then read a PR and press merge. “Not sure” is a complete answer to the middle one. No call, no screen share. 24 hours from a complete intake, not from the receipt.
What it is not: not a push to your default branch, not a zip of generic templates, and not hooks, MCP servers or 3–5 skills — that is Setup Sprint $99.
Because the free one does not know your repo. kit-claude-code-starter is a real, complete setup, and it is generic on purpose — you fill in your own commands, your own traps, your own skill. Setup Lite is that same shape written from your code, with every command run there first. If you would rather do that reading yourself, take the starter. It is free and we would rather you had it than nothing.
No. The intake form asks in plain language and “not sure” is an answer — working the rest out of your code is the job. You do need Claude Code installed to get anything out of the files, though: they are configuration that sits in your repo and does nothing on its own. If none of this is familiar yet, start with Claude Code for non-developers.
Lite is one repo, one skill, the allowlist, 24 hours. Sprint is 3–5 skills, hooks, up to 3 MCP servers wired and a one-page runbook, 48 hours. If you already know you want hooks or an MCP server, buy the $99 — Lite would only need upgrading, and a scope change is a new order.
Collaborator access to a branch, and nothing else. The deliverable is a pull request you review, never a push to your default branch. We never ask for an API key or a .env, and if you send one by accident we will tell you to rotate it and keep it out of the work.
Comment on the PR and we fix it against the scope — one round, included. Full refund if the PR is not open yet. After it is open we fix rather than refund; at $29 we are not going to argue about it.
Both, and we would rather say so. AI tools including Claude produce the work, and a person reads every deliverable before it ships. The commands-were-run check above is that person’s job.
Then this is not the right buy, and we will say so rather than stretch it. Take the free starter, build something with it, and come back when there is a repo to set up.
Most people should stop at Lite. These are the three questions that come back after a Lite PR is merged, answered the same way we would answer them in an email.
Only if the thing that is still annoying you is on this list: Claude repeating a mistake that a hook would have blocked, more than one repeated job that wants its own skill, or a tool it cannot reach — your issue tracker, your database, your docs — that wants an MCP server. Setup Sprint $99 is those three things plus a one-page runbook, in 48 hours. If what is still annoying you is none of those, Sprint will not fix it and you should not buy it.
No. Sprint is a separate $99 order, the same as it is for someone who never bought Lite — each package is its own scope and the terms say so. What carries over is the work: the CLAUDE.md from your Lite PR is the starting point for Sprint, not something written again, and your skill stays as one of the 3–5. You are paying for the hooks, the extra skills, the MCP wiring and the runbook, which is what the $99 is.
Yes, and a fair number of people should. You have a working CLAUDE.md and a skill to copy the shape of, the starter repo shows the settings and hooks layout, and the Skills & Hooks Pack $29 is 10 skills and 8 hooks to adapt rather than write. That is the cheaper path and we would rather point at it than sell you the same afternoon twice. Buy Sprint when you would rather not spend the afternoon, or when the MCP wiring is the part you do not want to debug.
.mcp.json stubs for the servers you wantThis one starts with the form, not a checkout. Two of these run at a time, so we confirm a slot and send a payment link before you pay anything. If both slots are full you hear that with a date, not a week later.
What you send: the repo URL, what it is built with, the install / dev / test / lint commands that work today, and the three workflows that annoy you most.
Scope, plainly: one repo per purchase, and a monorepo counts as one. It is configuration, not feature work — no bug fixes and no features in your codebase. You bring your own Claude subscription; nothing here resells or shares access. Never put a token or key in the form — we do not ask for one and we do not want one.
Smaller than this? Setup Lite $29 is one repo and one skill as a PR. Setup Sprint $99 adds hooks and up to 3 MCP servers wired. Want a tool built rather than configured? That is MCP Basic $199.
At a complete intake — the repo, the commands and the three workflows — not at the receipt. That is the honest version: we cannot write a CLAUDE.md for a repo we have not seen. Two of these run at a time, so the order is the other way round from the download packs: you send the form, we confirm a slot and send a payment link, then you pay. If both slots are full you hear that with a date, within a day, instead of paying into a queue.
Full refund if we have not started. Once the work is delivered we fix it against the scope rather than refund — one round, included. Because the slot is confirmed before you pay, “we have not started” is the usual state if you change your mind. The full wording is in the terms; if a case does not fit it cleanly, email support@sdvsignal.com and we will answer in plain words rather than quote a clause at you.
One repo per purchase — a monorepo counts as one, two products in two repos is two purchases, and it is better to hear that now than on day four. It is configuration, not feature work: no bug fixes, no features written in your codebase, no refactors. It is async — no call, no pairing session, no shared Slack channel. If what you actually need is one tool built rather than a repo configured, MCP Basic $199 is the honest purchase and we will say so.
Four things, by email: a CLAUDE.md tuned to your repo with real commands rather than a template, .mcp.json stubs for the servers you picked, a permissions note saying what to allow and what to keep behind a prompt, and a walkthrough video 20 minutes or less. A person reads all of it before it ships.
No key, ever. The .mcp.json stubs reference your environment variables — we do not ask for, hold or accept a token, and if one arrives in the form we tell you it came through in plaintext and to rotate it. You do need your own Claude subscription: nothing here provisions, resells or shares access.
Buying a download — the ASO Launch Pack DIY ($19) or the MCP Config Pack ($49)? You can skip this form. There is no scope to agree and no start date to pick; the link is emailed to you the same day, to the address on your Stripe receipt.