mcp-contract-security
# mcp-contract-security
MCP server for smart contract security scanning.
Implements the required tools:
- `scan_contract`
- `compare_security_models`
- `audit_checklist`
Supports languages:
- `solidity`
- `rust`
- `typescript`
## What it checks
`scan_contract` currently detects heuristic patterns for:
- Reentrancy
- Integer overflow/underflow risk
- Access control issues
- Front-running exposure
- Oracle manipulation risk
The output includes:
- vulnerability list with severity/type/line
- `security_score`
- `near_equivalent_score`
- NEAR-specific security notes and recommendation
## Install
```bash
npm install -g mcp-contract-security
```
Published package:
- npm: https://www.npmjs.com/package/mcp-contract-security
- MCP Registry: https://registry.modelcontextprotocol.io/v0/servers/io.github.mastrophot%2Fcontract-security-scanner/versions/0.1.1
## MCP config (Claude Desktop)
```json
{
"mcpServers": {
"contract-security": {
"command": "mcp-contract-security"
}
}
}
```
## Tool usage
### `scan_contract`
Input:
```json
{
"code": "contract source code here",
"language": "solidity"
}
```
### `compare_security_models`
Input (optional):
```json
{
"language": "solidity"
}
```
### `audit_checklist`
Input (optional):
```json
{
"language": "rust"
}
```
## Local development
```bash
npm install
npm run check
```
## Deliverable assets
Additional publish assets are prepared in `deliverables/`:
- `deliverables/mcp-registry-submission.md`
- `deliverables/security-subreddit-posts.md`
- `deliverables/blog-why-near-contracts-safer.md`
- `server.json` (MCP Registry metadata, schema-validated)
## License
MIT
TDQS
Scored across 3 tools
Each tool has a clearly distinct purpose: scanning contracts for vulnerabilities, comparing architectures, and providing checklists. There is no meaningful overlap that would cause an agent to select the wrong tool.
Two tools follow a verb_noun pattern (scan_contract, compare_security_models), while audit_checklist is more of a noun phrase. The pattern is mostly predictable and readable, with only a minor deviation.
Three tools is a reasonable scope for a focused contract security server. It is slightly small but each tool serves a distinct, justified function within the domain.
The toolset covers scanning, comparative analysis, and actionable guidance, which gives agents a solid workflow. It lacks deeper features like detailed report generation or code fixes, but the core security assessment use case is well covered.