solana.new: an AI-native CLI from idea to raise

solana.new is an AI-native builder toolkit for Solana from SendAI and Superteam. Where create-solana-dapp or anchor init hand you a template, solana.new is built around a coding agent (Claude) and a curated knowledge base — and it spans the whole arc from picking an idea to generating a pitch deck.

One command installs it into your agent environment:

curl -fsSL https://www.solana.new/setup.sh | bash

It works with Claude (the default), and also Cursor and Codex.

The four stages

• Idea — 500+ startup concepts curated from YC, Alliance, Superteam, and SendAI, to validate or pick a direction.
• Build — AI code generation for programs (with testing and audit passes) or full apps, wired to real protocol integrations.
• Launch — GTM help: marketing copy, X threads, launch strategy.
• Raise — competitive-landscape analysis and investor-grade pitch-deck generation.

What it actually installs: skills, repos, MCPs

The real substance is the curated catalog it brings into your agent's context — a library of Solana skills (task-specific playbooks), reference repos, and MCP servers. Instead of the model guessing from training data, it has vetted, current patterns to pull from. (If you're reading this on devrels.xyz: the skills powering this site's own Claude Code environment come from exactly this catalog.)

Integrations span the ecosystem it scaffolds against — Jupiter, Helius, Meteora, Phantom, Privy, Pump.fun, Orca, Birdeye, Sanctum, Kamino, SendAI, and Dflow — so generated code targets the tools teams actually use.

Why this shape matters

It's a concrete take on the "docs-as-context" idea we've seen elsewhere (the ProofNetwork MCP is a single-protocol version): the most useful thing you can give a coding agent isn't a starter template, it's authoritative, current playbooks and the MCP tools to use them. solana.new bundles that for the whole ecosystem and wraps it in a workflow.

The honest read

solana.new is opinionated and broad — that's the strength and the caveat. The idea/launch/raise stages are genuinely useful for a solo builder or hackathon team who'd otherwise stall on "what" and "how to pitch," but the output is a starting point: AI-generated programs still need real review and audits before mainnet (the toolkit says as much by including testing/audit passes). The catalog is only as good as its curation is fresh — its value is that SendAI + Superteam maintain it. If you already know your stack and your idea, you may only want the skills/MCP layer, not the full idea-to-raise funnel. As an on-ramp for new Solana builders, though, it's one of the most complete things out there.

References

• solana.new
• sendaifun/solana-new (GitHub)
• Solana AI agents — the broader landscape
• The docs-as-MCP pattern, single-protocol edition

The future of Solana onboarding probably isn't a template you clone — it's a curated brain you hand your agent. solana.new is a bet on exactly that.