Kestrova

Solidity security scans for AI coding agents (Slither). Free preview with no API key; agents pay per scan with x402 (USDC on Base or Solana): 0.25 USDC for every finding with file and line, 1 USDC for a full report with fixes, PDF and SARIF. Automated scan, not an audit.

Barındırılan MCP Sunucusu

npx add-mcp 'https://kestrova-api.onrender.com/mcp?src=registry'

Claude Code, Codex, Cursor ve daha fazlasına kurulur

Dokümantasyon

The same Slither scanner the website uses, available to scripts, CI and AI coding agents. The free preview needs no key. Paid results are per call with x402 on Solana or Base, signed by your own wallet tool, and only after you approve the price: every finding with file and line for 0.25 USDC (agent tier, JSON), or the full report with a plain-English explanation and fix for every finding plus PDF and SARIF for 1 USDC.

Check a deployed contract by address

Send a contract address instead of source code, on Base, Ethereum, BNB Chain, Arbitrum, Optimism or Polygon ("chain": "ethereum", …; default base). Kestrova reads the verified source from Sourcify or the chain's Etherscan explorer, follows proxies to the current implementation and checks only the files the contract uses. Use it before your agent buys, approves or interacts with a contract.

curl -s https://kestrova-api.onrender.com/preview/address \
  -H 'content-type: application/json' \
  -d '{"chain": "base", "address": "0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913"}'
  • POST /preview/address or MCP scan_contract_address_preview: free counts, detector names and one example finding, plus the contract name, compiler and proxy.
  • POST /scan/address or MCP scan_contract_address: optional, 0.03 USDC: every finding with file and line, JSON.
  • Unverified contracts are never scanned or charged; the free check returns hints read from the bytecode instead (likely mint, pause, blacklist, fee or upgrade powers).
  • Results are cached until the on-chain code changes, so an upgraded proxy is always checked again.

Automated pattern detection, not an audit. It does not detect honeypots, hidden taxes or rug pulls.

REST API

Free preview: severity counts and a description of each High/Medium issue (issues: what and why, without file, line or fix), plus one lower-severity finding in full as an example (example_finding). Exact lines are in the paid findings result; fixes are in the full report. Package imports such as @openzeppelin/contracts/... stay as written: send the package files under that path, under node_modules/, or in the Foundry folder lib/openzeppelin-contracts/contracts/. Up to 15 KB / 10 files, 10 calls per hour. The paid full report (POST /scan) covers up to 60 KB / 20 files.

Each free preview includes behind_paywall: exactly what the paid report would add for that code, and it says so when the report would add nothing. After a paid POST /scan, the X-Kestrova-Report-Token response header can be exchanged at POST /scan/report-file with {"token": "…", "format": "pdf" | "sarif" | "json"} for the PDF, SARIF 2.1.0 or JSON report. An unpaid POST /scan answers 402 with the terms both in the PAYMENT-REQUIRED header and in the JSON body.

Paying from code: use any x402 v2 client. Check the terms first (the free MCP tool payment_terms, or the 402 body), sign exactly one accepts[] entry, and repeat the request with the PAYMENT-SIGNATURE header. Pay only with your user's approval; after an unclear result, recover instead of paying again.

curl -s https://kestrova-api.onrender.com/preview \
  -H 'content-type: application/json' \
  -d '{"sources": {"Token.sol": "pragma solidity ^0.8.20; contract T {}"}}'

POST /scan returns every finding with locations. Without a payment it answers 402 Payment Required with the terms in the PAYMENT-REQUIRED header and runs nothing. Use an x402 client to pay; you are charged only after a successful scan.

Response fields: findings and counts cover every finding; total_counts equals counts. Each finding has a scope: application (your code, including findings that also touch dependencies) or dependency (located only in uploaded third-party code under lib/<package>/, node_modules/ or @scope/package/). application_counts and dependency_counts split the counts; dependency_findings repeats the dependency ones for a separate section. Dependency findings keep their original severity and are not confirmed false positives.

Interactive API reference (OpenAPI) · openapi.json

Remote MCP server

Streamable HTTP endpoint for any MCP client:

https://kestrova-api.onrender.com/mcp
  • scan_solidity_preview: free counts and detector names
  • scan_solidity_findings: optional, 0.25 USDC: every finding with file and line, JSON (x402 payment in _meta["x402/payment"])
  • scan_solidity_full: optional, 1 USDC: the same findings plus an explanation and fix for each, and PDF/SARIF report files
  • payment_terms: current price and payment details

Claude Code Tested

Install the plugin (skill and MCP server together):

npm install -g kestrova-scan
claude plugin marketplace add "$(npm root -g)/kestrova-scan"
claude plugin install solidity-security-review@kestrova

Or add only the MCP server:

claude mcp add --transport http kestrova "https://kestrova-api.onrender.com/mcp?src=skill"

Codex, Cursor, VS Code / Copilot, Gemini CLI

Tested Gemini CLI connects to the MCP server. Setup instructions For Codex we checked the generated configuration only; Cursor and VS Code follow their documentation and were not tested by us.

OpenAI Codex (CLI and IDE):

codex mcp add kestrova --url "https://kestrova-api.onrender.com/mcp?src=codex"

Gemini CLI:

gemini mcp add --transport http kestrova "https://kestrova-api.onrender.com/mcp?src=gemini"

One-click links open Cursor or VS Code and ask you to confirm the server. Or add it by hand:

Cursor, in ~/.cursor/mcp.json:

{"mcpServers": {"kestrova": {"url": "https://kestrova-api.onrender.com/mcp?src=cursor"}}}

VS Code / GitHub Copilot, in .vscode/mcp.json:

{"servers": {"kestrova": {"type": "http", "url": "https://kestrova-api.onrender.com/mcp?src=vscode"}}}

Agent-readable summary of everything on this page: /llms.txt.

Claude.ai and ChatGPT

Setup instructions Kestrova is a remote MCP server with no sign-in, so it can be added as a custom connector. Menu names can differ by plan and app version.

Claude.ai: Settings → Connectors → Add custom connector. Name: Kestrova, URL:

https://kestrova-api.onrender.com/mcp?src=claude-ai

ChatGPT: Settings → Apps & Connectors → Advanced settings → turn on Developer mode, then create a connector with the URL below and authentication “No authentication”.

https://kestrova-api.onrender.com/mcp?src=chatgpt

Then ask, for example: “Check this Solidity contract with Kestrova.” The free preview runs without a wallet; the paid tools only return their price unless your client has an x402 wallet and you approve the payment.

Paying as an autonomous agent

  • Price: 1.00 USDC per full report (POST /scan, 1000000 base units) or 0.25 USDC per findings result (POST /scan/findings, 250000 base units), x402 v2 exact, Solana or Base mainnet only. You are charged only if the scan succeeds; recovery works the same for both.
  • Wallet (Solana): the signer needs the price in USDC and no SOL; the PayAI facilitator pays the network fee. The signer signs one USDC transferChecked; it never broadcasts.
  • Never pay twice: after pending, already_paid or a timeout, recover the report with a wallet-signed challenge at POST /scan/recover. Recovery is free and never a payment.
  • Base: also accepted as USDC on Base mainnet (x402 v2 exact, EIP-3009 transferWithAuthorization): the signer needs 1.00 USDC on Base and no ETH; the facilitator submits the authorization. Recovery uses an EIP-191 signature from the paying address.

AI agent skills

For agents that support SKILL.md skills. The skill tells the agent to run the free check after writing Solidity, to explain results without overclaiming, and never to pay without your explicit approval.

The skill ships in the npm package (skills/solidity-security-review); npx kestrova-scan --setup claude (or cursor, vscode, codex, gemini) prints the MCP configuration for your agent.

npm package

CLI, agent skill and Claude Code plugin (MIT licensed; contains no server code): npmjs.com/package/kestrova-scan

GitHub Action

Free check of the changed Solidity files on every pull request. Only new findings are reported, in the job summary; fail-on: high blocks a pull request that adds a new High finding. It needs no token or secrets, never pays, and only warns if the service is busy. GitHub Marketplace: Kestrova Solidity Scan

on: pull_request
permissions:
  contents: read
jobs:
  kestrova:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
        with: { fetch-depth: 0, persist-credentials: false, submodules: recursive }
      - uses: actions/setup-node@v4
        with: { node-version: 22 }
      - uses: kestrova/solidity-scan-action@v1.0.0
        with: { fail-on: high }