v0.8.5 · alpha · MIT

One idea in.
Every platform out.
In your voice.

Open-source content automation. You write one anchor. Springy cascades it into native, on-brand drafts for LinkedIn, X, Threads, Bluesky, Reels, TikTok, Medium, and more. Local-first, Claude or GPT, human-gated. Not a SaaS.

Get started See how it works

$ npm install -g @andersonfda/springy

anderson@laptop ~/my-brand — springy demo

$ npx springy demo

── [1/5] RANKED OPPORTUNITY ──────────────

rank 1 · score 87/100

title Anthropic ships Claude Sonnet 5 — "deep reasoning" goes public

── [2/5] CASCADE GRAPH ──────────────────

        RANKED OPPORTUNITY
              │
              ▼
         ANCHOR ESSAY  (critique loop ×3)
              │
       ╔══════╧══════╗    ← you approve
       ║   HITL #1   ║
       ╚══════╤══════╝
      ┌───┬───┼───┬───┐
      ▼   ▼   ▼   ▼   ▼
   LinkedIn  Thread  Bluesky  Reel  TikTok  …
              │
       ╔══════╧══════╗    ← you approve
       ║   HITL #2   ║
       ╚══════╤══════╝
              ▼
       PUBLISH QUEUE (API → browser fallback)

── [3/5] ANCHOR ESSAY ───────────────────

# I tried Claude Sonnet 5 for 4 hours. Here's what changed.

[…truncated — full essay ~1200 words, critiqued against voice-RAG]

── [5/5] COST ESTIMATE ──────────────────

Anchor essay (critique ×3) ~$0.18

5 platforms × critique loop ~$0.40

Humanize + images ~$0.40

Total ~$0.98

$ ▊

`npx springy demo` — offline, zero cost, just run it

Not a SaaS. Not an agent. A graph.

Three things make springy output sound like you, not like another ChatGPT Substack post.

Brand-voice RAG

Every generation retrieves your 5 best past posts (embedded with Voyage-3) and conditions on them. You're extending your own voice, not fighting GPT-voice.

Critique loop

Generator → critic (8 dimensions, 9 for long-form with a vulnerability check) → revise → humanize. Hard gates prevent obvious AI slop from shipping.

Two human gates

One on the anchor essay. One on the fan-out. You approve what goes in the queue. Publishing is explicit: --live plus a one-time "type YES" marker per project.

Different from Buffer. Different from Zapier.

Schedulers don't write. Agents wander. Springy is the third shape.

	Buffer / Typefully	Zapier / Make	Raw LLM agent	springy
Writes the post	you do	optionally (generic)	yes, but drifts	yes, in your voice
Multi-platform fan-out	manual per channel	you glue it	whatever it feels like	deterministic graph
Local-first data	SaaS cloud	SaaS cloud	cloud	your laptop
Human approval gates	scheduling only	if you build them	rarely	two, always
Extensible	no	yes, via their DSL	prompts-as-config	plugins + MIT
Total cost	$15–$30 / mo	$30–$100 / mo	your API bill	$5–$30 / mo (API only)

Who this is for.

Indie hackers and solopreneurs building an audience alongside a product and tired of rewriting the same idea six times.

Agencies and small teams running multiple brands who need consistent voice without hiring a content manager.

Technical founders who'd rather npm install than sign up for a SaaS. CLI-first, configurable, forkable.

If you want a drag-and-drop visual editor and zero Terminal time, springy is not for you yet.

60 seconds to try.

Zero API keys. Zero cost. Just see what it does.

~/anywhere

# Run the offline demo — zero keys, zero cost.
$ git clone https://github.com/andersonfda/springy
$ cd springy
$ npm install
$ node bin/cli.js demo

# When you like what you see:
$ node bin/cli.js doctor       # preflight
$ node bin/cli.js cascade --top # your first real run

Full quickstart Star on GitHub

$ ./ship-with-confidence.sh

Distribution is a tax on builders.

Stop paying it with your afternoon. Write the thing. Let the pipeline do the translation.

Install in 5 min See what it costs

MIT · built in public · github.com/andersonfda/springy

One idea in. Every platform out. In your voice.