DFlow & Phantom: Our Claude Code Skill for Solana Apps

AI can generate almost any function you ask for, but it still struggles with architecture.

Full-stack crypto apps are not just interfaces or isolated APIs. They are coordinated systems where wallets, identity, execution, and settlement operate together in real time.

From our conversations with Solana developers, the struggle is rarely about writing code. It’s about wiring systems together correctly. Most teams end up stitching together examples from multiple repos, and debugging edge cases that stem from integration gaps, not logic errors.

That’s why today, we’ve teamed up with Phantom to release our Claude Code skill.

The DFlow & Phantom Claude Code skill gives your agent the missing structure needed to build full-stack Solana applications, with the wallet and liquidity layers designed as a single system from a single prompt.

Download the skill:

https://github.com/DFlowProtocol/dflow_phantom-connect-skill

Why This Matters

If you’re building a consumer crypto app, two layers must work together seamlessly:

1. Wallet layer → authentication, signing, session management
2. Liquidity layer → swaps, prediction markets, price feeds, settlement

This skill teaches Claude how to build full stack trading apps for spot crypto and prediction markets. Everything from the front-end wallet connection to the backend trading infrastructure can be built with a single prompt.

The Wallet Layer: Phantom Connect

Through Phantom Connect, your agent understands how to:

• Support browser wallet connections
• Enable embedded wallets via social login
• Deep link into the Phantom mobile app
• Scaffold React and React Native wallet flows correctly
• Handle authentication, signing, and network configuration cleanly

The wallet flow is now structured from the start.

The Liquidity Layer: DFlow

DFlow provides the execution infrastructure behind the applications your agent builds.

Today, that infrastructure supports over 1 million monthly active traders, with $46.5B in swap volume since April 2025 and $22M in prediction market volume since December 2025.

On the execution side, DFlow powers:

• Token swaps
• Prediction market trading
• Real-time pricing
• End-to-end settlement patterns

Instead of improvising, the skill encodes actual integration patterns used in production.

Your agent learns:

• How wallet state flows into trade execution
• How to call DFlow endpoints correctly
• How to structure UI updates around fills
• How to gate prediction markets behind identity verification
• How to scaffold a complete trading interface

The result: a working system, not disconnected code blocks.

What We’ve Already Built With It

Internally, we’ve used the skill to:

• One-shot a functional DEX UI
• Build a fully working prediction market website in hours
• Scaffold a React Native app with Phantom Connect + DFlow trading
• Add Proof-based identity verification before enabling prediction market trading

All without bouncing between tabs or stitching together old tutorials.

How This Connects to Our MCP Server

Last week we launched DFlow MCP.

https://x.com/dflow/status/2024547541807767757

The MCP server gives agents direct access to live DFlow documentation at generation time.

This skill does something different:

• DFlow MCP → real-time reference
• Claude Code skill → repeatable implementation workflows

DFlow MCP provides authoritative source access.

The Claude Code skill provides structured execution patterns.

Together, they replace scattered search with structured shipping.

Download the skill:

https://github.com/DFlowProtocol/dflow_phantom-connect-skill

Set up DFlow MCP:

https://pond.dflow.net/build/mcp