Roadmap

Coral is built in five stages. Each stage targets a distinct audience and unlocks the next. We commit to things publicly because companies that do are trusted more.

V1 Shipping late May 2026

The Wedge

Target: Late May 2026 — currently in build sprint

Audience: Claude Code power users (developers).

Memory engine, skills marketplace, agent loadouts, creator upload, 3-tier pricing, web app only.

  • Memory engine: client-side encrypted, captures every Claude Code conversation automatically
  • Recall that feels smart — injects the 3–5 memories that matter at the start of every session
  • 5 launch loadouts covering backend, frontend, AI building, automation, and research
  • 50+ hand-curated skills at launch, growing weekly
  • Creator upload live from day one — publish your own skills and lock in the Founding Creator flag for a permanent +5% revenue share bonus once payouts ship in V1.5
  • Stripe billing, 3-tier pricing (Free / Pro $12 / Pro+ $29)
  • Installable web app — add to home screen on any device
  • Trust strategy: open memory protocol, real founder bio, public changelog + roadmap
V1.5 Next — 4-6 weeks after V1

Creator Payouts + Model-Agnostic Backend

Target: 4-6 weeks after V1 ships

Founding Creators start earning money. The first creator-paid marketplace in the Claude ecosystem. Plus the memory classifier becomes model-agnostic — Claude stays the default, but GLM-5.1 and any OpenAI-compatible model become first-class options.

  • Stripe Connect integration — creators get paid monthly
  • 20% platform fee — matching Raycast, the only comparable platform for dev tools
  • Creators set their own prices — or keep skills free
  • License enforcement — paid skills require a license key to install
  • Founding Creator bonus — permanent +5% revenue share lift on any skill uploaded before V1.5 goes live
  • Tax forms auto-generated (1099-K for US creators, VAT for EU)
  • Disputes, refunds, and content moderation handled by Coral
  • Model-agnostic classifier backend — pick Claude (default) or GLM-5.1 in settings. Any OpenAI-compatible provider supported. Your memory stays identical either way.
  • Launch partner for GLM-5.1 memory support — complete with a dedicated loadout
V2 Planned — 8-12 weeks after V1

Coral: Memory Worlds

Target: 8-12 weeks after V1 ships

Audience: Devs (carryover) + prosumers (power users, content creators, students, knowledge workers).

The marquee feature: a 3D universe of your memories. Zoom from a solar system all the way down to walkable ground. Categories become planets. Memories become places you can visit. Runs in any modern browser — no plugins, no installs. The first AI memory tool you'll actually want to explore.

  • Continuous zoom from the full memory universe down to individual memories
  • Coral orb hovers above the worlds as the living center of the experience
  • Classic flat-list view stays available for anyone who prefers it
  • Coral Desktop installer (Mac + Windows + Linux) — no terminal required
  • 30–50 more curated marketplace skills
  • 5–10 more loadouts
V2.5 Planned — 4-8 weeks after V2

Coral 1.5: Worlds Expansion

Target: 4-8 weeks after V2 ships

Two big visual additions in one launch event:

  • Themed environments per category — each planet gets its own personality and vibe, so exploring your memory feels like walking through distinct places instead of scrolling a list
  • Agents as living lights — when your Coral agents are working, you can see them moving through your memory in real time. Activity made visible.
V3 Planned — 3-4 months post-V1

The Bridge

Target: 3-4 months after V1 launch

Audience: Devs sharing Coral with non-technical friends, family, co-workers.

  • Coral Control mobile app — manage agents from your phone, push notifications, simple chat interface
  • Coral as an AI persona — friendly character handles permission requests and API setup
  • Mobile screen-context feature — on-demand only, not always-watching. Cover the screen, ask for help, get context-aware advice.
  • Native video watching — point Coral at any YouTube link and she analyzes it for you (transcripts, visual understanding, summaries). Out of the box.
  • Auto-skill installation with consent
  • Quota awareness — Coral knows your Claude plan and warns before tasks blow your monthly budget
V4 Planned — 6-9 months post-V1

Mass Market

Target: 6-9 months after V1 launch

Audience: Non-technical users discovering Coral via word-of-mouth.

  • Coral Cloud Run — scheduled jobs that keep working when your computer is off, with full access to your encrypted memory context. Complements Anthropic's Managed Agents rather than competing with them. Usage-based pricing on top of Pro+.
  • Coral Voice — talk to your agents from your phone or desktop
  • Onboarding for users who don't know what Claude is — assumes zero tech vocabulary
  • Simplified templates that hide all dev language
V5 Planned — 9-12 months post-V1

Coral Business (separate product)

Target: 9-12 months after V1 launch

Audience: Existing businesses adopting AI for the first time.

Ships as a separate product with two tiers once Coral has built brand recognition through V1-V4. The waitlist on /business collects emails throughout that bridge period.

  • Coral Business — solo business owners ($75-150/mo)
  • Coral Business Pro — teams of 2+ business operators ($150-300/mo or $50-75/seat). Multi-user collaboration with mutual approval gates is the headline.
  • Guided 30-minute setup with a real human reviewing the config
  • Custom agent team for each business's specific operations
  • Mobile-first management ("run your AI workforce from your phone, like Slack")
  • Dedicated support tier
V5 Spin-off — launches with V5

Coral Together (consumer spin-off)

Target: launches alongside Coral Business

Audience: Non-technical people building creative things together.

Two (or more) friends, one shared project, each with their own AI agent team. A "business meeting" mode where both humans and their AI teammates plan together. Mutual approval on anything that spends money or touches the outside world.

  • Business meeting mode — human + human + Coral in one shared conversation
  • Visual roadmap with task self-assignment to each person
  • Mutual approval gates — spending money requires both humans to say yes
  • Separate agent teams per user — you keep your workflow, they keep theirs
  • Target use cases: presentations, hobby projects, wedding planning, video games, content creation
  • Pitch: "Build anything with anyone. The AI co-founder for the rest of us."