Netlify Free for static sites and Functions: local tests, credits and the signup blocker

What we tested and how to reproduce it

The fixture contains one HTML file, one JavaScript Function and a netlify.toml. The Function uses the Web Request/Response interface: GET returns a small JSON object, while POST echoes a JSON body. The browser screenshot comes from the real running fixture. No vendor account, payment method or hosted site was needed for these local checks. Netlify documents this local workflow in its CLI guide and the runtime interface in its Functions documentation.

netlify dev --offline --no-open --skip-gitignore   --dir /absolute/path/to/fixture/public   --functions /absolute/path/to/fixture/functions   --port 5683

We used explicit absolute paths because this fixture lived inside the sudoai.net repository. Our first run loaded the parent repository’s go Function instead of the fixture’s echo Function. Its rewritten endpoint returned 404. Supplying both paths loaded echo and fixed the problem. This is a monorepo/configuration discovery observation in our run; we did not establish a general CLI defect.

The measurement: 50 requests through Netlify Dev

We made one unreported warmup GET, then 50 sequential GETs to /api/echo using Python Requests. Each sample includes local HTTP transport and the development proxy/function process. There was no load generator, intentional parallelism or internet hop. Tests ran on this Linux box; its physical location and provider capacity were not independently verified. An Asia/Singapore timezone does not prove Singapore network egress.

The relatively large local latency is a reason to measure the deployed route before deciding, not evidence that Netlify’s CDN is slow. Development emulation and production execution have different processes and scheduling. We retained both the initial failed routing run in logs and the successful functional checks; the table reports only the final corrected 200-response sample.

Free-plan limits that affect an indie release schedule

These values are documented, not limits we reached. Prices and limits were checked Oct 1, 2026. New accounts have used credit-based plans since September 4, 2025; eligible older accounts can retain legacy plans. Switching to credits is irreversible according to the plan documentation. Read your own dashboard before applying this table to an existing account.

All meter rates above come from how credits work; concurrency and members come from the plan comparison. When the credit balance is exhausted, the documentation says all web projects on that balance pause. We did not deliberately exhaust credits and cannot show the pause page from our own account.

Three budget examples, calculated rather than measured

These scenarios are arithmetic using the published rates, not invoices or observed quotas. Actual metering granularity and the dashboard’s rounding were not tested. AI inference is another shared-budget expense; this review did not invoke models, Agent Runners, a database or any paid add-on. Keeping an unused AI feature out of a hosting estimate is sensible, but enabling it later requires a fresh budget.

What user reports add—and what they do not prove

A saved Hacker News search includes a developer reporting that frequent scheduled rebuilds burned their credit allocation and that they moved the build schedule to GitHub Actions (original comment). Another commenter describes adding credits after HN traffic increased demand (original comment). These are individual reports, not our reproduction. The actionable point agrees with the official shared-budget model: publishing cadence and traffic compete for the same allowance.

GitHub issue #8423 reports Functions failing when netlify dev --cwd is invoked outside a project. We did not reproduce that specific error; our problem was resolving the Function folder. The distinction matters when troubleshooting. A Reddit credit-pause discussion was found, but direct retrieval returned 403, so it is not used as factual evidence here.

Netlify or Cloudflare for this workflow?

If your app mainly serves static files and makes a few API requests, read our Cloudflare Pages hands-on review beside this one. That article includes an actual hosted deploy; this article cannot offer an equivalent Netlify hosting benchmark. Our site runs on Cloudflare Pages, so migrating merely to repeat static hosting would need a concrete preview, integration or operational benefit.

Netlify’s named deployment contexts and Function routing are useful when an existing team workflow is already built around its configuration. For a new small site, first estimate publications, transfer and request volume. Using an external build runner does not by itself remove Netlify’s production-publication charge. Fewer builds, fewer production publishes and smaller assets address different meters.

Do not confuse ordinary Functions with Edge Functions. They have different execution environments; code and dependency compatibility should be tested in the environment you will actually deploy. Our fixture is a regular Function. We did not validate Edge Functions, provider regions, function timeouts, production concurrency or cold starts.

The paid step and migration checks

The current public plans list Personal at $9/month with 1,000 credits and Pro starting at $20/month with 3,000 credits and unlimited members. Pro has additional credit tiers. Those are subscription prices, not all-in quotes: optional recharge and add-ons can add cost. Every paid tier is not tested. The April 2026 pricing announcement and changelog help explain why older articles about per-seat pricing may disagree.

Before moving production, make a preview deploy and test redirects, security headers, environment-variable contexts, asset caching and the real function URL. Test a failing Function and a rollback as well as a happy-path request. Confirm whether production becomes public only after publication in your account’s current workflow. Finally, alert on the shared credit balance and identify who can publish. None of these hosted checks was completed in this run.

What would change our verdict?

A real free account with a successful deploy, measured publication time, before/after credit screenshots and external latency would let us assess hosting rather than tooling. Until then, our verdict is limited: Netlify Dev is usable for this tiny static-plus-Function fixture after setting explicit paths; the documented credit model needs a release budget. We will not rank its production performance from localhost numbers.