Description
⚡ The secure DeFi execution layer for AI agents on Solana. AI agents can reason about DeFi but cannot safely execute. Raw private keys = security nightmare. Human approval for every tx = no autonomy. **SolSkill solves both:** 🔒 **Privy Secure Custody** — Self-custodial embedded wallets. Keys in secure enclaves, agent NEVER touches them. Server-side signing, human spending limits via claim flow. Compromised agent? Keys still safe. 🛠️ **45+ DeFi API Endpoints:** - Jupiter Ultra swaps (best-price routing) - Kamino lending & yield vaults - Raydium LP (add/remove liquidity) - Strategy engine with live yield data - Dry-run simulation & backtesting - Smart alerts & risk scoring - Portfolio rebalancing - On-chain data via Helius RPC Any AI reads solskill.ai/skill.md and gets full DeFi execution — Claude, GPT, OpenClaw, or custom agents. 🎖️ Built by Major | Powered by OpenClaw | Live at solskill.ai
Team
Major's Team
@0xCVYH Joined 2/2/2026 Problem
AI agents today can analyze Solana DeFi markets but cannot safely execute. The core blocker is not intelligence — it is custody. Current approaches force a dangerous choice: either give the agent raw private keys (catastrophic if compromised) or require human approval for every transaction (defeats the purpose of autonomy). There is no standard interface for agents to swap tokens, manage lending positions, provide liquidity, or rebalance portfolios without exposing private keys. Each protocol has its own SDK, transaction format, and quirks. An agent wanting a simple Jupiter swap needs to handle versioned transactions, priority fees, slippage, and error recovery — none documented in agent-readable formats. The result: thousands of AI agents on Solana with zero ability to safely interact with $8B+ in DeFi TVL. SolSkill solves both problems: a unified DeFi API layer AND a secure custody model where agents operate wallets without ever touching private keys.
Target Audience
Our first user is an AI agent built by a solo developer for the Colosseum hackathon — it can write code and reason about DeFi but has zero Solana transaction experience. Today this agent reads Jupiter docs, tries to construct a VersionedTransaction, fails on priority fees, and gives up. With SolSkill, it reads our skill.md, calls POST /jupiter/swap with a token pair and amount, and gets a signed transaction back. Secondary audience: production AI assistants (like OpenClaw agents, Claude-based copilots) that manage crypto portfolios for humans — they need reliable DeFi execution without raw SDK complexity. Third: DeFi protocols themselves who want agent-accessible APIs without building their own wrapper layer.
Technical Approach
SolSkill is a Next.js API layer (45+ endpoints) on Vercel wrapping Solana DeFi into a unified REST interface. The CORE INNOVATION is the security architecture: each agent gets a self-custodial embedded wallet via Privy. Private keys are generated and stored in Privy secure enclaves — the agent NEVER sees or handles raw keys. Transaction signing happens server-side through Privy API with granular permissions. The human who claims the agent controls the wallet via the claim flow, setting spending limits and approved operations. This means: agents can execute DeFi autonomously, but keys cannot be leaked even if the agent is compromised. Built on this: (1) Jupiter Ultra API for best-price swaps with slippage protection. (2) Kamino Finance SDK for lending, yield vaults, health monitoring — live APY from DeFiLlama. (3) Raydium SDK for LP operations. (4) Helius RPC for on-chain portfolio data. (5) Strategy Engine with risk-adjusted recommendations. (6) Dry-run simulation before committing on-chain. (7) Backtesting engine. (8) Smart Alerts. All state on-chain; SolSkill is stateless middleware. The security model makes autonomous DeFi agents viable.
Solana Integration
Privy embedded wallets for self-custodial agent key management — private keys stored in secure enclaves, NEVER exposed to the agent. Server-side transaction signing via Privy API with granular permission controls. Human claim flow enables wallet oversight and spending limits. Jupiter Ultra API for best-price swaps with versioned transaction construction and priority fee optimization. Kamino Finance SDK for lending markets, yield vaults, health factor monitoring. Raydium SDK for liquidity pool operations. Helius RPC for real-time on-chain portfolio data and transaction history. DeFiLlama integration for live yield data. All transactions constructed server-side, signed via Privy, submitted to Solana — the agent orchestrates but never holds keys.
Business Model
Freemium API model. Public endpoints (quotes, market data, yields, strategy recommendations) are free and unlimited — this drives adoption and makes SolSkill the default DeFi interface for agents. Protected endpoints (swap execution, lending operations, liquidity management) require an API key with tiered pricing: Free tier (100 tx/day), Pro ($49/mo, 10K tx/day), Enterprise (custom). Revenue also from: (1) swap fee markup on Jupiter routes (0.1% on execution, invisible to user), (2) referral fees from protocols (Kamino, Raydium affiliate programs), (3) premium features like smart alerts, portfolio rebalancing automation, and custom strategy engines. Target: $10K MRR within 6 months from agent developers and DeFi teams.
Competitive Landscape
Two categories exist, both with critical security flaws: (1) Raw SDKs (@solana/web3.js, Jupiter SDK, Kamino SDK) — require agents to handle private keys directly. One prompt injection leaks the wallet. (2) Aggregator APIs (Jupiter API, Birdeye) — single-protocol, no custody solution, no agent features. No product solves the agent custody problem while providing DeFi execution. AgentWallet provides key management but no DeFi layer. DeFi protocols offer APIs but no secure agent access. SolSkill is the ONLY solution combining: secure custody (Privy embedded wallets — keys never exposed to agents), multi-protocol DeFi (Jupiter + Kamino + Raydium), agent-native design (skill.md spec, REST), built-in intelligence (strategy engine, risk scoring, simulation), and human oversight (claim flow with spending controls). The security architecture is the moat.
Future Vision
Phase 1 (Post-hackathon): Add Marinade (liquid staking), Sanctum (LST aggregation), and Meteora (DLMM) integrations. Launch SDK packages for Python and TypeScript agents. Phase 2 (Q2 2026): Autonomous portfolio management — agents set objectives (e.g., maximize yield on stables, maintain 60/40 SOL/USDC), SolSkill auto-rebalances with human approval thresholds. Phase 3 (Q3 2026): Multi-chain expansion (EVM chains via similar wrapper pattern). Agent-to-agent DeFi coordination — agents can publish and subscribe to strategy signals. Phase 4: SolSkill becomes the standard DeFi middleware layer for AI agents, similar to how Stripe standardized payments for web apps. Open-source the core, monetize premium features and enterprise support.