Shovl
Shovl
Pricing

Plans that match a personal library

Target per-workspace quotas below are sized from public Supabase and OpenAI list pricing (April 2026). Final checkout may tune numbers slightly; the structure stays: free to try, paid for storage + AI load, no ads on your saves.

Starter

$0/ forever

Enough to run the full loop while staying inside a typical free-tier footprint.

  • 500 MB total file storage (PDFs, images, uploads)
  • 35 new captures / month (note, URL, or file)
  • 25 grounded library chat replies / month (answers with citations)
  • Per-file cap 25 MB
  • Hybrid search + shared background job queue
Get started
Likely default paid tier

Plus

$12/ per month

Sized for weekly saves; quotas align with Supabase Pro–class headroom and steady OpenAI usage.

  • 12 GB total file storage
  • 120 new captures / month
  • 200 grounded library chat replies / month
  • Per-file cap 75 MB
  • Higher job priority when the worker queue backs up
Join waitlist

Library

$22/ per month

Heavy personal archives: big PDF stacks, frequent chat, room to grow before you hit platform ceilings.

  • 40 GB total file storage
  • 320 new captures / month
  • 650 grounded library chat replies / month
  • Per-file cap 150 MB
  • Best-effort fastest ingestion; early access to Chrome + mobile capture when shipped
Join waitlist

Cost basis (list prices, April 2026)

Shovl's stack in PLAN.md: Supabase Auth, Postgres + pgvector for chunks and hybrid retrieval, Storage for raw files, OpenAI for text-embedding-3-small embeddings and gpt-5.4-mini for titles, summaries, tags, and grounded chat. Rounded figures for planning:

  • OpenAI: text-embedding-3-small ≈ $0.02 / 1M input tokens; gpt-5.4-mini (standard) ≈ $0.75 / 1M input, $4.50 / 1M output (cached input lower).
  • Supabase: Free tier often includes on the order of ~500 MB Postgres per project and ~1 GB file storage (metered as GB-hours)—see database size and storage pricing. Pro (≈ $25 / mo list) typically includes much larger file storage (on the order of 100 GB bundled) and a larger included database footprint (on the order of 8 GB) before overages—exact bundles change with Supabase; check the live page.
  • pgvector: embeddings live in Postgres; at ~1.5k dims and personal library sizes, vector tables stay a small slice of the DB versus chunk text and FTS—so file + chat caps usually dominate cost before vector index upgrades matter.

How we're thinking about monetization

PLAN.md keeps the product narrow: one private workspace per person, Postgres + worker ingestion, hybrid search, chat that cites retrieved chunks. Pricing mirrors vendor math instead of opaque “enterprise seats.”

Costs follow usage

Enrichment and chat use OpenAI (e.g. gpt-5.4-mini); chunk embeddings use text-embedding-3-small. Heavier months mean more tokens—paid tiers widen caps instead of ads or data resale.

Storage + Postgres

Original files sit in object storage; chunk text, tsvectors, and pgvector embeddings sit in Postgres. Plans scale stored files and keep vector growth personal-scale (exact search on pgvector per PLAN.md).

Chat packs for spikes

If you blow past monthly replies, optional top-up packs can cover bursts without jumping tiers—same idea as in Terms.

Annual and founding options

Yearly billing with a discount rewards commitment; founding pricing can help fund extension and mobile without per-seat enterprise pricing.

Questions? ma10016@nyu.edu