Dev tools · checked 2026-10-19

Can I vibecode Social Fetch?

NOT REALLY· don't bother

Social Fetch puts one key and one JSON shape in front of public data from TikTok, Instagram, LinkedIn, X, YouTube and more than a dozen other sites, fetched live on each request. The code is the easy part. Instagram, TikTok, LinkedIn and Facebook work to stop automated reads, so what you pay for is proxies, sessions and parsers that someone repairs when a site changes. An agent can build a clean fetcher in an evening for the platforms with open APIs, such as YouTube, Reddit, Bluesky, GitHub and Hacker News; keeping the hostile ones working is an ongoing job.

$29/moprice

$348it costs you a year

weekendbuild time

—people replaced it

Before you start

What you need on hand.

DifficultyIntermediate · Some system setup or keys.

TimeAn evening for the platforms with open APIs

Claude Code, Codex or CursorAny of them can run the prompt.
Python 3.12A virtual environment keeps the dependencies tidy.
A YouTube Data API keyCreate a project in Google Cloud and enable the YouTube Data API v3.
A Reddit script app and a GitHub tokenBoth are free and take a few minutes each.

Before you buildSocial Fetch gives 100 credits on sign-up and 250 more once you save a card, and a profile lookup is usually 1 credit. Try your real queries there before you build anything.

1. Install one agent: Claude Code, Codex, Cursor.
2. Make an empty folder and open it in the agent.
3. Paste the prompt below, press enter, and answer its questions as it builds.

The prompt

Build me a personal social-data fetcher to replace Social Fetch for the platforms that allow it. Requirements:

- Python 3.12: FastAPI served by uvicorn on http://127.0.0.1:4900 and a `typer` CLI
  (`sfetch youtube-video <url>`, `sfetch reddit-search "<query>"`); both call the same functions.
- One module per provider behind a common interface: YouTube (Data API v3 with YOUTUBE_API_KEY
  for channels, videos and comments, plus `youtube-transcript-api` for captions), Reddit (`praw`
  with a script app: subreddit posts, threads, search), Bluesky (public XRPC endpoints, no key:
  profile, author feed, post thread), GitHub (REST with GITHUB_TOKEN: users, repos, issues) and
  Hacker News (the Algolia search API).
- One envelope for every response: {status: found | not_found | private | error, platform, kind,
  data, fetchedAt}. Normalise id, url, author, text, createdAt and counts the same way in every
  provider and keep the provider's raw payload under `raw`.
- Storage: SQLite in ./data/cache.db with a TTL per kind (default 10 minutes), a log of every
  call (provider, kind, status, milliseconds) and a daily quota counter per provider that refuses
  a call once the limit set in .env is reached.
- Errors: retry 429 and 5xx up to 3 times with exponential backoff and honour Retry-After; never
  retry other 4xx; return the provider's message in the envelope.
- Tests: `pytest` with recorded responses (`vcrpy`) so the suite runs without any key.
- Single user. No accounts, no telemetry, local-first; keys live in .env and the only outbound
  calls go to the providers above.
- Out of scope: Instagram, TikTok, LinkedIn, Facebook and X (they block unauthenticated reads and
  need rotating proxies and logged-in sessions), media downloads, and proxies of any kind.
- README: how to get each key, the quota per provider, and a plain warning that reading a platform
  against its terms is your risk.

Opens your agent with the prompt filled in, you press enter. Or copy it and paste it into an empty folder.

Share on X

Max MiskowSponsoredAI agents, automations and MVPs, built for your company.Visit site ›

Then keep going

Next prompts, same folder.

The first prompt builds the core. Run these one at a time in the same folder to grow it into your own Social Fetch.

Add account watching to the fetcher in this folder. Requirements:

- A `watch.yml` file listing accounts per provider (YouTube channels, Bluesky handles,
  Reddit users, GitHub users).
- A command `sfetch watch run` that fetches each entry, stores a snapshot in SQLite and prints
  new posts and changes in follower or star counts since the last run.
- Respect the daily quota: stop and report which entries were skipped.
- An optional `--markdown` flag that writes the report to ./reports/YYYY-MM-DD.md.
- Update the README with an example cron line.

Add a yt-dlp provider to the fetcher in this folder. Requirements:

- Call `yt-dlp --dump-json --skip-download` through a subprocess with a 60-second timeout;
  never download the media itself.
- Map the output into the shared envelope (title, author, description, duration, view and like
  counts, upload date) and keep the full JSON under `raw`.
- A `subtitles` kind that returns captions as plain text with timestamps.
- If yt-dlp is missing or the site is unsupported, return status `error` with a clear message.
- Update the README with the install command and a note to keep yt-dlp up to date.

Add CSV export to the fetcher in this folder. Requirements:

- A command `sfetch export --kind <kind> --since <date> --out file.csv` that writes cached
  results of one kind as flat columns (id, url, author, text, createdAt, counts).
- Flatten nested fields with a dot in the column name; always quote text and escape newlines.
- Skip rows whose status is not `found` unless `--all` is given.
- Update the README with two examples.

What you lose

Live reads of Instagram, TikTok, LinkedIn, Facebook and X without running proxies or logins
The same JSON shape for 20+ platforms, including comments, transcripts and hosted media files
Search, ad libraries and follower lists on platforms that hide them
A vendor that repairs adapters when a platform changes
A published uptime page and a request ID on every response for support

Why people still pay

Teams building trend trackers, research tools and AI features need data from sites that don't want to be read, and running their own proxies and scrapers is a second job. A few hundred lookups a month fit a $14 credit pack with no subscription, and the monthly plans scale the price per credit down as volume grows.

Moatinfrastructure scale integrations

Do the math

Social Fetch or your own copy?

Over

Your usage

Count my time12h ×$/h

Social Fetch, 1 yr$348

Your own copy, 1 yr$35running + build session

You keep$313pays for itself in the first month

Estimate · The official APIs are free at personal volume and the cache runs on your machine. The saving ends where you would need proxies for the hostile platforms, which is the reason for the verdict. The build is token usage on your agent plan.

Leaving Social Fetch

Take your data, then cancel.

1 · Take your data out of Social Fetch

1Social Fetch fetches live data on each request, and the docs describe no stored library, so there is nothing to export: keep the JSON your code already saved.
2Check what is left before you leave: GET /v1/balance is free and returns your credit balance.
3Each response carries a requestId; save the ones you may need for a support or refund question.

2 · Cancel Social Fetch

1Cancel at any time from the billing page.
2To use the 30-day refund, ask support within 30 days of your first plan payment: it is a full refund, or the unused part if you used more than 10% of the included credits. The plan ends and its included credits are withdrawn, while pay-as-you-go credits remain.
3The refund applies once per person.

Pay-as-you-go and refill-pack credits never expire, so a leftover pack is worth keeping as a fallback.
Completed lookups are charged even when the result is not_found, and cache hits are charged at the full price: cache on your side before you migrate.
Replace calls platform by platform: official APIs for YouTube, Reddit, Bluesky and GitHub first, and keep Social Fetch for the hostile platforms until you decide whether you need them.
Read each platform's terms before you scrape it yourself.

Checked 2026-10-19 against Social Fetch's docs: 1, 2, 3

The escape hatch

Don't feel like building it? These are already free.

yt-dlpOpen sourceUnlicense downloader and metadata extractor for YouTube, TikTok and a long list of other video sites; subtitles included.★ 199k · active Oct 2026 · one-click install
CLIVisit ↗InstaloaderOpen sourceMIT Python tool for Instagram profiles, posts, stories and hashtags; works best with your own logged-in session and can be rate-limited.★ 14k · active Oct 2026 · one-click install
CLIVisit ↗gallery-dlOpen sourceGPL-2.0 command-line downloader for image and media galleries across dozens of sites, with metadata output as JSON.★ 20k · active Oct 2026 · one-click install
CLIVisit ↗

All 6 free alternatives to Social Fetch ›

The numbers

Social Fetch pricing

Plan	Monthly	Annual, per month	What you get
Free credits	$0	$0	100 credits on sign-up, plus 250 if you save a card before buying anything.
Pay as you go	custom	—	One-time packs: 1,000 credits for $14, 25,000 for $59, 230,000 for $379. One person, no subscription.
Starter	$29/workspace	$25.83/workspace	18,000 credits a month, 1 seat.
Growth	$79/workspace	$70.83/workspace	50,000 credits a month, up to 5 seats.
Scale	$225/workspace	$202.50/workspace	150,000 credits a month, up to 25 seats (more on request).
Enterprise	custom	—	For 1M+ credits: invoicing and custom terms.

Free tier: A one-off 100 credits, or 350 with a saved card; there is no monthly free allowance.
Billing: Monthly or annual subscriptions plus separate pay-as-you-go packs. Cancel any time from billing. A plan's price is locked while you stay subscribed. Refund: ask support within 30 days of your first plan payment for a full refund, or for the unused part if you used over 10% of the included credits; once per person.
Hidden costs: A credit is not a request: a profile lookup is typically 1 credit, but transcripts, hosted media files and audits cost more (a transcript is 1 credit plus 10 per started minute for AI fallback). Lookups that end in not_found, private or restricted are still charged, and cache hits cost full price. Add-ons: priority support $19, shared Slack $49, custom roles $49, SSO and directory sync $149 a month.
Source: www.socialfetch.dev

Also on the list

Bolt.new$25/mo · Dev toolsKINDA

Cursor$8/mo · Dev toolsKINDA

GitHub Copilot$10/mo · Dev toolsKINDA

Questions

Social Fetch, briefly

Not answered here? Tell us and we'll add it.

Ask a question ›

Sign-up gives 100 credits with no card; saving a card adds 250 more, and nothing is charged until you buy.

Not really. Teams building trend trackers, research tools and AI features need data from sites that don't want to be read, and running their own proxies and scrapers is a second job. A few hundred lookups a month fit a $14 credit pack with no subscription, and the monthly plans scale the price per credit down as volume grows. The prompt on this page builds the closest honest substitute.

About $29 a month on the Starter plan, $348 a year (checked 2026-10-19).

Live reads of Instagram, TikTok, LinkedIn, Facebook and X without running proxies or logins; The same JSON shape for 20+ platforms, including comments, transcripts and hosted media files; Search, ad libraries and follower lists on platforms that hide them; A vendor that repairs adapters when a platform changes; A published uptime page and a request ID on every response for support. If any of those matter to you, keep paying.