solanashield-mcp
v2.0.0
Published
AI-powered smart contract security MCP server for Solana — x402 pay-per-audit, 12 tools, 40 vulnerability patterns
Maintainers
Readme
SolanaShield MCP
AI-powered smart contract security for Solana programs via Model Context Protocol
SolanaShield is a specialized MCP server that brings 40+ Solana vulnerability patterns to any AI assistant. It detects critical security issues in Anchor/Solana programs including missing owner checks, arbitrary CPI, PDA confusion, integer overflow, reentrancy, and more.
Built for the Colosseum Frontier Hackathon (Apr 6 — May 11, 2026).
Features
- 40 Vulnerability Patterns — 8 Critical, 12 High, 12 Medium, 8 Low
- 12 MCP Tools — Full audit, targeted checks, risk scoring, fix generation
- Zero Dependencies — Pure static analysis, no external APIs needed
- Solana-Native — Every pattern is specific to Solana/Anchor programs
- JSON Output — Machine-readable results for CI/CD integration
Quick Start
npm install -g solanashield-mcpOr clone and build:
git clone https://github.com/ElromEvedElElyon/solanashield-mcp.git
cd solanashield-mcp
npm install && npm run buildClaude Desktop Configuration
Add to ~/.config/claude/claude_desktop_config.json:
{
"mcpServers": {
"solanashield": {
"command": "node",
"args": ["/path/to/solanashield-mcp/dist/index.js"]
}
}
}MCP Tools
| Tool | Description |
|------|-------------|
| audit-program | Full security audit with severity-sorted findings |
| check-accounts | Account validation (owner, signer, authority) |
| check-cpi | Cross-Program Invocation safety analysis |
| check-pda | PDA seed confusion and bump storage checks |
| check-arithmetic | Integer overflow/underflow and precision loss |
| check-tokens | SPL Token authority, mint, and freeze checks |
| risk-score | Quick 0-100 risk score for triage |
| list-patterns | Browse all vulnerability patterns |
| explain-finding | Detailed exploit scenarios and remediation |
| generate-fix | Auto-generate code fixes for findings |
| compare-versions | Diff two code versions for regression analysis |
| audit-json | Structured JSON output for programmatic use |
Vulnerability Categories
| Category | Patterns | Example | |----------|----------|---------| | Access Control | 3 | Missing owner check, missing signer | | CPI Safety | 4 | Arbitrary CPI, reentrancy, privilege escalation | | PDA Safety | 2 | Seed confusion, bump not stored | | Token Safety | 4 | Authority mismatch, mint/freeze authority | | Arithmetic | 2 | Integer overflow, precision loss | | State Safety | 5 | Reinitialization, account close, rent exemption | | DeFi | 1 | Flash loan via CPI | | Oracle Safety | 1 | Stale price feed | | Deserialization | 2 | Type cosplay, unsafe deserialization | | Anchor | 2 | Missing constraints, init_if_needed | | DoS | 1 | Unbounded iteration | | Logic | 1 | Timestamp dependence | | Observability | 2 | Missing events, missing logging | | Optimization | 2 | Unused accounts, CU optimization | | Code Quality | 3 | Magic numbers, test code, missing docs | | Configuration | 1 | Hardcoded addresses | | Error Handling | 1 | Missing error context | | Maintenance | 1 | Deprecated API usage |
Example Output
Risk Score: 62/100 (HIGH)
Critical: 1 | High: 3 | Medium: 1 | Low: 3
Total findings: 8 across 19 lines
RECOMMENDATION: Do NOT deploy without fixing critical and high findings.Architecture
src/
index.ts # MCP server with 12 tool registrations
patterns/
solana-patterns.ts # 40 vulnerability pattern definitions
analyzers/
static-analyzer.ts # Pattern matching engine + report formattingLicense
MIT
Author
PadraoBitcoin — [email protected]
