Manifestack
A free, read-only skill for Claude Code, Cursor and Codex. Ask your agent “what will we pay at 50k users?” and get each bill ahead of you with the fix that cuts it. Plan limits, outages and compliance gaps come in the same report.
npx manifestack --agent claude-codeThen run /manifestack in your project.
Claude Code~/relaylog
> we're aiming for 50k users by Q3. what will that cost us?✓ Reading STACK.md EU data only, budget ~$600/mo✓ Scanning the repo 7 services ✕ Vercel auto-pause at $200 takes the site down✕ Resend weekly digest hits the rate limit✕ Supabase database is in the US, rule says EU! Clerk second login system next to Supabase ? CDN requests last month? not in the code; Vercel → Usage › 6.1M ✓ Read 21 vendor pages, priced 50k usersAt 50k users: ~$4,580/mo on a ~$600 budgetBreaks first: Vercel, ~18th of every month10 findings, each with a fix; 3 take ~15 min today Fix today, ~15 min:→ Vercel Billing → Flat Rate CDN, saves ~$340/mo→ OpenAI $100 in credits lifts the cap to $5,000/mo→ Resend move to the 100k tier before NovemberSaves ~$325/mo now; at 50k: ~$4,580 → ~$1,920/mo ✓ Updated .manifestack/STACK.md> 100+ vendors, with AWS, Google Cloud and Azure service by service. Prices read live from their own pages; any other vendor works too.Full list
The bill
~$310 vs ~$25
A Kubernetes cluster, a multi-zone database, a NAT gateway and a load balancer for 40 users. One app host and managed Postgres would carry them. Right at scale; not at forty users.
Built for 1M users~$310/mo
Right-sized for 40~$25/mo
Same app, same 40 users~$3,400/yr extra
Read-only
Supabase Free goes read-only at 500 MB. At 2 MB a day that's months away, until a launch or an untrimmed log table adds 365 MB in three weeks. No signups, no orders.
500 MB
On the +2 MB/day trend: week 33Actual: week 11
EU vs us-east-1
Your customers are in the EU, and your terms say their data stays there. The database went up in us-east-1 for a weekend prototype and never moved. The first security review will find it.
Run /manifestack or just ask your agent. Unlike a search or a chatbot, it knows your repo and your numbers, reads today's vendor pages, and remembers every decision.
Say what you're building, for whom and on what budget. Get a whole stack, each pick explained and priced on today's vendor pages, from launch to 100k users.
| 1k users | $20/mo |
|---|---|
| 10k users | $65/mo |
| 100k users | ~$420/mo |
what should we build this on?
Point it at any repo, yours or a client's. It finds every service across frontend, backend and infra, and flags what will cost too much, hit a limit or break a rule.
| web/ | Vercel, Clerk, Supabase |
|---|---|
| api/ | OpenAI, Stripe, Supabase |
| worker/ | Resend, Sentry |
7 services across 3 apps, checked against STACK.md
why is our bill growing?
Both sides priced on your workload and checked against your code, including what a switch would really take.
| Supabase, now | Neon | |
|---|---|---|
| covers | database, auth, files | database, auth, files |
| at 9k users | Free now, $25/mo on Pro | ~$20/mo, usage-based |
| in your code | supabase.* in 14 files | 14 files to rewrite |
| switch takes | EU region move, ~1 day | ~3 days, users and files too |
Stay and move to an EU region. Neon covers the same jobs at about the same price, so switching adds ~2 days of rewrites for no gain at 9k users.
should we move to Neon?
You ask the agent to add login, and it installs Clerk. Before it finishes, it tells you:
+ "@clerk/nextjs": "^6"
Supabase Auth already signs users in, so this makes two login systems, and Clerk bills per user. Keep one?
Runs on its own in Claude Code and Cursor.
Output of /manifestack audit on a sample product: Next.js, Python and Go, 9,000 monthly users, seven services. The verdict comes first: what today's fixes save and what 50k users will cost. Below are 7 of its 10 findings, each with how bad it is, when it hits, what it costs and the fix.
~/relaylog/manifestack auditSample output
CriticalRisk
Overage ~$340/mo; Spend Management set to $200 with Pause Production Deployments on
CriticalRequirement
Project in us-east-1, STACK.md requires an EU database
fra1.CriticalRisk
worker/main.go sends ~9,000 digests with 50 in parallel and drops the error
HighBill
3.4 TB last month, 1 TB included, 6.1M CDN requests
HighRisk
$310/mo, +21% a month
MediumLimit
312 / 500 MB, +1.1 MB/day
events rows older than 90 days, or take Pro with the EU move.LowOverlap
@clerk/nextjs and supabase.auth both in web/src/
Usage figures come from numbers you give it, exports in the repo or a read-only vendor MCP. A report is as current as those inputs.
## Requirements
budget: ~$600/mo
users: 9k now, 50k by Q3
requires: EU database, SOC 2 vendors
avoid: Kubernetes # no ops on-call
## Database: Supabase
plan: free
usage: 312 MB, +1.1 MB/day
source: supabase.com/pricing # read 2026-10-06
decided: stay on Free until first paying user
revisit_when: db_size > 400 MB
next: Pro, $25/mo
Requirements, plans and the reason behind each choice live in .manifestack/STACK.md, in your repo. The next chat, the next teammate and every new service start from it.
The highlighted line: once the database passes 400 MB, the next audit says it's time for Pro.
It runs inside your agent, with no Manifestack server. It never asks for API keys, .env values or logins, and its only network calls read vendors' public pages.
Developers, small teams and agencies who build with a coding agent and pick their own services (Vercel, Supabase, Stripe, AWS and the like), from the first commit to a few hundred thousand users. If nobody on the team owns reading pricing pages, it does that for you. It is not a FinOps tool for large cloud accounts.
A plain agent answers from training data that can be a year old, then forgets. Manifestack reads today's vendor pages, prices your actual usage, and remembers your requirements in STACK.md, so every later change is checked against them.
From each vendor's pricing and limits pages, read during the run, not from training data. It prices your own usage: numbers it finds in the repo, ones you give it, or a vendor's read-only MCP. Every figure cites the page and the date it was read, and a page it can't open is marked unverified instead of guessed.
No. Bills are where most teams feel it first, so the report leads with them. The same audit flags plan limits that take a site down, data rules such as EU residency or SOC 2, two services doing one job, and setups built for far more users than you have. Each finding has a when, a cost and a fix.
No. It is read-only. It writes only inside .manifestack/: STACK.md, plus working files git ignores. Decisions go into STACK.md only after your yes.
No. It reads your repo and env var names, never their values, and asks you for the usage numbers it can't see. If you've connected a vendor's read-only MCP (Supabase, Neon, PostHog), it can read usage there instead.
Yes: Claude Code, Cursor, Codex, GitHub Copilot, Windsurf, OpenCode and any agent that reads Agent Skills. npx manifestack finds them for you. The automatic check on every new service runs in Claude Code and Cursor; elsewhere, run /manifestack audit before merging a new service.
Yes. It reads JavaScript, Python and Go dependencies, config files and env var names, plus infrastructure code (Terraform, CloudFormation and SAM, Bicep, Pulumi) and CI deploy steps. React Native and Expo apps are covered, including RevenueCat, Adapty and EAS. For native iOS and Android it asks about the services it can't see.
Yes. MIT license, no paid tier, no account. There is no Manifestack server to pay for.
Free and open source. Read-only, no API keys, no account.
npx manifestack --agent claude-codeThen run /manifestack in your project.