Sandcastles – AI Short-Form Video Research, Hooks & Scripts

Description

Sandcastles is an AI research and scriptwriting workspace for short-form video. It helps marketers and creators find top channels on Instagram Reels, TikTok and YouTube Shorts, surface videos that beat each creator’s own baseline (“outliers”), pull transcripts and breakdowns, then generate hooks and scripts from those patterns. It does not film, edit or publish the video — that stays in your usual production stack.

The product is built around a credit system (video analysis, new scripts and reports each cost credits), with Pro and above including an MCP connector so Claude, ChatGPT, Cursor and other assistants can query the same niche data. The vendor claims 100k+ users and a large outlier database; treat those as vendor claims. Co-founder Kane Kallaway publicly introduced the current Sandcastles wave around September 2025, with Collections, Reports and MCP shipping through 2026.

Please note: this listing covers sandcastles.ai (short-form research and scripts). It is not the Vectorpark “Sandcastles” beach game on iOS, idle sand-castle mobile games, or Sand Castles Studio.

Who It’s For

• Social and content marketers who post Reels, TikToks or Shorts several times a week and already spend hours studying competitors.
• Founder-led brands and personal-brand operators who need a repeatable idea → hook → script pipeline without hiring a full research desk.
• Agencies and multi-client teams that want separate workspaces, verified client analytics and (on Titan) guest access plus API.
• Teams already writing in Claude or ChatGPT who want live niche performance data via MCP instead of generic “viral hook” advice.

It is a weaker fit for casual posters, pure long-form YouTube SEO (Sandcastles focuses on Shorts-style content, not long-form imports), or anyone who mainly needs an AI video editor or auto-clipper.

Core Capabilities

Watchlists and outlier feeds. Build a feed of Instagram, TikTok and YouTube channels in your niche, then filter videos by outlier score, views, engagement, keywords, platform and recency. Outlier scoring compares a clip with that creator’s usual performance, not raw celebrity view counts alone.

Analysis to script. Analysing a video spends one credit and unlocks transcript, idea analysis, hook, storytelling format and visual layout. From there you can shortlist ideas, save hook formats, export for another LLM, or spend another credit to write a new script. Edits to an existing script (rewrite, translate, length change) are free.

Automations, reports and personal analytics (Pro+). Rules can auto-analyse videos that hit your outlier/view/engagement thresholds. Reports include Content Strategy Audits and Hook Genius-style outputs. Verifying your own channels unlocks a dashboard with refresh intervals that tighten as you move up plans.

Collections and MCP. 100+ curated collections (formats, editing styles, visual hooks, signature series) are browseable without credits on every plan. MCP on Pro, Visionary and Titan connects the same research into Claude, ChatGPT, Codex, Cursor, Grok and other MCP clients; asking about already-analysed data is free, new analyses still cost a credit.

Key Features Summary

• Cross-platform watchlists: Instagram, TikTok, YouTube Shorts
• Outlier detection against each creator’s baseline
• Transcripts plus hook / storytelling / visual breakdowns
• Hook vault and scriptwriter (remix, from scratch, or fix an existing script)
• Ideas inbox, shortlist, projects and collections
• Automations, bulk analyse, saved filters, Persona and exports on Pro+
• Reports (including Content Strategy Audit) and verified-channel personal analytics
• MCP connector (Pro+); REST API and workspace guests (Titan)

Plans & Pricing

Source: sandcastles.ai homepage pricing and help centre plan comparison (updated 2 October 2026), checked 6 November 2026. Credits: 1 per video analysis or new script; reports cost a base amount plus up to 1 credit per newly analysed video. Unused plan credits do not roll over (monthly or annual). Paid subscription plans are non-refundable per FAQ; Starter is the exception on cancel. Homepage CTAs say “Try It Risk-Free” / “Try it free” but help docs state Starter is a paid $5 plan, not a free trial. Confirm live prices before buying.

Pros & Cons

Pros

• Clear research-to-script workflow across Reels, TikTok and Shorts in one feed
• Outlier scoring versus each creator’s baseline is more useful than raw view sorting alone
• Published credit rules and plan matrix; $5 Starter is a low-risk way in for new accounts
• MCP on Pro+ lets marketers keep research inside Claude/ChatGPT/Cursor

Cons

• Pre-production only — no editing, captions, voiceover or scheduling
• 100 Pro credits can feel tight if you analyse several references per post
• Advertised “from $39/mo” is annual billing; monthly Pro is $49; no refunds on paid plans
• Homepage “4.92/5” and “100k+ users” are not tied to a named review platform; no clear Product Hunt listing found

Where It Fits

Sandcastles sits with social media marketing, content marketing and video marketing tools that solve ideation and scripting. Pair it with an AI video editor or clipper for production, and a scheduler for publishing. It is not a GEO/AEO tracker, not a long-form YouTube SEO suite, and not an influencer CRM.

FAQs

What is Sandcastles?

An AI platform for short-form video research and scriptwriting: watchlists, outlier analysis, hooks and scripts for Instagram, TikTok and YouTube Shorts.

How much does Sandcastles cost?

New accounts can start on Starter at $5 a month. Full plans: Pro $49/$39, Visionary $99/$79, Titan $499/$399 (monthly vs yearly-equivalent monthly). Homepage FAQ says plans start at $39/month when billed annually on Pro.

Is there a free trial?

Help docs say Starter is a paid $5 plan, not a free trial. Cancelling Starter ends access immediately and refunds the latest $5. Older testimonials still mention a free trial; treat that as outdated.

Does Sandcastles edit or publish videos?

No. It researches, analyses and scripts. Filming, editing and posting happen elsewhere.