Resources

Roadmap

A public roadmap so you know where we're headed. No firm dates (the project is a solo effort) but the order of the priorities reflects what we're working on.

Want to influence the orderIf a feature further down the list is blocking you, tell us by email at contact@forgent.me. The concrete needs of real users always take priority over theoretical ideas.

Recently shipped (July 2026)

The on-chain pivot: identity, rewards, and the flagship model. See the changelog and the token doc for the details.

  • Sign-In With Solana: your wallet is your account (Phantom, Solflare, Backpack…). No email, no password — Google login is retired.
  • Free Builder and deploy: designing, deploying, and running an agent need no API key — everything is platform-funded. There is nothing to plug in or configure.
  • $FORGENT creator points: publishing (+50) and completed public runs (+2) accrue points, convertible at token launch, with favorites and on-chain proof links on the marketplace.
  • Claude Fable 5: the flagship model is now selectable in the Builder and filterable on the marketplace.

Recently shipped (May 2026)

May 2026 marathon: full Advanced Mode + marketplace overhaul + infra hardening. 16 phases shipped, the product is now in a beta-ready state.

  • ✅ Direct editing + {{x}} variables + Dry-run & A/B testing + Versions & logs + Webhooks & workflows
  • ✅ Complete marketplace: 2-choice landing, dedicated Templates/Skills feeds, search bar, URL-stateful filter sidebar, spotlight, dynamic OG images
  • ✅ Publish flow: 3-state modal (claim username → type/category/slug form → result with copyable URL), 7 canonical categories
  • ✅ Fork DB with a transactional counter + owner notification + lineage snapshot that survives the deletion of the parent
  • ✅ Creator profile: hero with avatar + bio + 4 stat blocks, grid of published agents, enumeration resistance
  • ✅ Notifications inbox: desktop + mobile bell, 60s poll, mark-all-read on click
  • ✅ My Agents 3 tabs Created / Installed / Workflows with lineage chips and a published badge
  • ✅ Public stats(installs/forks/views) on a published agent's page + on the creator profile, denormalized with client-side sessionStorage dedup and server-side IP rate limiting
  • ✅ Production-ready infra: versioned Prisma migrations, Sentry integrated (no-op without a DSN), Upstash rate limiting on 5 critical endpoints, Playwright + GitHub Actions CI

Short term — In progress

  • Prod activation of Sentry + Upstash: the infra is wired, we just need to create the accounts (free tier is enough to start) and add the Vercel env vars.
  • Marketplace moderation V1: a flag/report flow for problematic published agents. The under_review state is already reserved in the schema.
  • Email notifications: in addition to the bell, send an email to the creator when their agent is forked (with opt-out in Settings).
  • Marketplace cold-start: seed 30-50 quality agents before the public opening so the marketplace already feels "alive" to the first visitor.

Medium term

  • In-place editing of a deployed agent: change the prompt, tools, or model of an existing agent without having to recreate it or redeploy it manually.
  • Custom UI for MCP servers: a wizard to register a custom MCP server directly from Forgent.
  • Cron-triggered workflows: V1 is ephemeral on the browser side, V2 will store and run in the background with an email notification at the end.
  • Paid premium Templates: the DB schema is already ready (priceCents Int?). Still missing the checkout flow + creator payouts.
  • Redis cache for the marketplace corpus: the current 10s in-memory cache doesn't survive Vercel cold starts. Migration to Upstash KV for stable latency.

Long term

  • Optional Pro plan: for those who want the advanced features (Advanced Mode, premium Skills, team collaboration, analytics). The core of the product (simple builder, deployment, running your agents) will stay free forever.
  • Persistent memory & success criteria: reintroduced in a simpler form. Anthropic's Managed Agents support memory stores and graders; we'll expose them once we have a UI design that makes them understandable for non-technical users.
  • Open source: publishing the code once the stack has stabilized.

Not planned

For transparency, here's what we will not do:

  • Support for non-Anthropic models (OpenAI, Mistral, Llama). The product is and will stay focused on Claude.
  • Custom hosting of agents (outside of Managed Agents). Anthropic handles that better than we would.
  • Selling your data or your usage to third parties. Never.

For features that have actually shipped, see the Changelog.