ai.greyridgesignals/tech-risk
Free MCP security scan: tool-poisoning & prompt-injection audit (OWASP LLM01/08) + trust badge
Installation
Ajoutez ceci à la configuration MCP de votre client (Claude Desktop : claude_desktop_config.json ; Cursor : .cursor/mcp.json ; VS Code et les autres utilisent la même forme « mcpServers »).
claude mcp add --transport http tech-risk 'https://api.greyridgesignals.ai/mcp'
{
"mcpServers": {
"tech-risk": {
"type": "http",
"url": "https://api.greyridgesignals.ai/mcp"
}
}
}
Documentation
Tirée du README du projet sur GitHub, actualisée chaque semaine. Les liens ouvrent GitHub.
Sommaire
- Grey Ridge Signals — Agent-Native x402 Data & Security APIs
- Two flagships — what agents actually pay for
- Endpoints
- MCP server (the pull channel)
- MCP client (mcp-client/ — the stdio companion)
- Call it (HTTP)
- Measured status
- How it runs
- License
- Agent-framework plugins (published)
- Python / LangChain / CrewAI
- Try it on YOUR token, free
- CI gate for MCP servers
Grey Ridge Signals — Agent-Native x402 Data & Security APIs
Pay-per-call crypto, DeFi, prediction-market, and on-chain data for AI agents — plus MCP security scanning — over the x402 payment protocol on Base mainnet.
No account, no API key, no subscription. An agent pays inline in USDC (from $0.001/call) and gets the data back in the same request. Every data endpoint has a free preview so an agent can taste the payload before paying.
- Base URL:
https://x402-data-api.sigrunner.workers.dev - MCP endpoint (streamable-http):
https://x402-data-api.sigrunner.workers.dev/mcp - Canonical domain:
https://x402-data-api.sigrunner.workers.dev(same API) - Discovery:
/.well-known/x402·/openapi.json·/llms.txt - Network: Base mainnet (
eip155:8453), USDC. Settles via the non-custodial xpay facilitator — no Coinbase/CDP account needed. Also listed on the CDP x402 Bazaar.
Two flagships — what agents actually pay for
- MCP server security scan (
/scan/mcp, $0.05) — a prompt-injection / tool-poisoning audit of an MCP server's tools (OWASP LLM01/LLM08). Zero-egress: for private/OAuth-gated servers you POST only thetools/listJSON — nothing else leaves your side. There is no free equivalent for this (unlike token-safety, where free multi-chain scanners exist). Free preview:POST /scan/mcp/preview. CI gate (on the GitHub Marketplace):uses: rezearcher/greyridge-x402-tools@v0.3.2. - Pre-trade decision gate (
/gate/swap, $0.008 ·/chain/token-report, $0.01) — a decision (GO/HOLD/BLOCK, SAFE/CAUTION/AVOID) sized to your trade, not raw flags your agent has to interpret. Base-native; if you trade other chains, keep your multi-chain scanner and use this on Base.
Everything below is the full catalog (crypto/DeFi data, on-chain reads, LLM, enrichment). Measured reliability at /status (not an SLA).
Endpoints
Crypto & DeFi data
| Endpoint | Price | Returns |
|---|---|---|
GET /crypto/prices?coins=bitcoin,ethereum,solana | $0.001 | Spot token prices (DefiLlama), keyless. Up to 25 CoinGecko ids. |
GET /crypto/funding?limit=20 | $0.001 | Cross-venue perp funding rates (Hyperliquid + OKX, graceful fallback) with the funding spread + best long/short venue — the arb signal, not just a single-venue list. |
GET /defi/yields?limit=20&project=&chain=&stable= | $0.001 | Top DeFi lending/LP yields (DefiLlama) with APY breakdown, 7d/30d APY trend, IL risk, and DefiLlama's stability forecast — decision-grade, not a raw list. |
GET /pm/markets?query=&limit=20 | $0.005 | Live Polymarket prediction markets: question, outcomes, prices, volume, liquidity, end date. |
Base on-chain (JSON-RPC read primitives)
Base-native, multi-provider RPC with automatic failover (never fails a paid call on one provider's rate limit).
| Endpoint | Price | Returns |
|---|---|---|
GET /chain/block-number | $0.001 | Current Base block number |
GET /chain/gas-price | $0.001 | Current gas price (wei + gwei) |
GET /chain/balance?address= | $0.001 | Native ETH balance |
GET /chain/token-balance?address=&token= | $0.001 | ERC-20 token balance |
GET /chain/tx?hash= | $0.001 | Transaction details |
GET /chain/receipt?hash= | $0.001 | Transaction receipt (status, gas used, logs) |
GET /chain/code?address= | $0.001 | Contract-code check — is it a contract, and EIP-7702 delegated-EOA detection (is_eip7702_delegate / delegate_address) |
GET /chain/wallet?address= | $0.003 | Bundle: ETH balance + tx count + contract flag in one call, EIP-7702 delegated-EOA aware (is_eip7702_delegate / delegate_address) |
Agent trading gates (new)
| Endpoint | Price | Returns | |
|---|---|---|---|
| `GET /gate/swap?token=0x…&side=buy\ | sell¬ional_usd=N` | $0.008 | Pre-trade go/hold/block decision for any Base ERC-20: safety verdict, risk score, pool liquidity and price-impact of your size, flags. One call before an agent swaps. Free live sample: /gate/swap/preview |
GET /chain/token-report?token=0x… | see /openapi.json | Full token due-diligence report (security + liquidity + metadata) in one call |
Security
| Endpoint | Price | Returns |
|---|---|---|
GET /scan/mcp?url= | $0.05 | Security audit of a target MCP server: tool-poisoning / prompt-injection / exfiltration / dangerous-capability / hidden-unicode (OWASP LLM01/LLM08) + risk score |
POST /scan/mcp {"tools":[…]} | $0.05 | Zero-egress audit — paste your own tools/list array; same itemized findings, nothing leaves your side and we never fetch/call your server. For OAuth-gated or private MCP servers that won't expose a URL |
GET /enrich/tech-risk?domain= | $0.05 | Tech-stack fingerprint → CVE (NVD) + EPSS + CISA-KEV attack-surface risk |
GET /enrich/domain?domain= | $0.01 | Firmographic + tech-stack enrichment, incl. full subdomain enumeration (subdomains[]) via crt.sh certificate-transparency logs, RDAP, DoH, HTTP fingerprint |
Free previews: GET /crypto/prices/preview, /crypto/funding/preview, /defi/yields/preview, /chain/block-number/preview, /chain/gas-price/preview (full live data, not truncated), /scan/mcp/preview?url=. Zero-egress MCP scan (for private/OAuth-gated servers): POST /scan/mcp/preview with {"tools":[…]} is the free counts-only preview; POST /scan/mcp with the same body is the paid itemized scan — nothing leaves your side. One-liner: examples/mcp-scan-preview.sh.
MCP server (the pull channel)
Add one remote MCP server and your agent gets the whole toolset. initialize and tools/list
are free so agents can discover the tools; paid tools/call returns an x402 challenge.
tools/list advertises 22 tools: crypto_prices, crypto_funding, defi_yields, pm_markets,
chain_block_number, chain_gas_price, chain_balance, chain_token_balance, chain_tx,
chain_wallet, chain_token_security (paid Base RPC + data, EIP-7702 delegated-EOA aware),
crypto_prices_preview, crypto_funding_preview, defi_yields_preview, pm_markets_preview,
chain_block_number_preview, chain_gas_price_preview, chain_token_security_preview (free),
and scan_mcp_server, scan_mcp_preview, enrich_tech_risk, enrich_domain (security).
// add to your MCP client config
{
"mcpServers": {
"grey-ridge-x402": {
"type": "streamable-http",
"url": "https://x402-data-api.sigrunner.workers.dev/mcp"
}
}
}
MCP client (mcp-client/ — the stdio companion)
For stdio-only MCP clients (Claude Desktop, Cursor, CLI agents), the repo ships a
companion package, x402-data-api-mcp, in mcp-client/. It runs on
the agent's own machine, speaks MCP over stdio, and proxies to the same Worker
/mcp endpoint — it adds no data surface, just a transport + x402 payment shim so
stdio-only clients can reach the same tools (22 tools, including paid ones when a
wallet key is configured).
Install / run (local path). The package is not yet published to npm, so run it from the repo:
cd mcp-client
npm install
npm run build # tsc → dist/server.js
node dist/server.js # start (or: npm start)
Published path (one line, no clone): npx @grey-ridge/x402-data-api-mcp — the package is
live on npm; its bin target is
dist/server.js, so npx runs the same server.
Configuration. Copy mcp-client/.env.example to
.env in mcp-client/: WORKER_BASE_URL defaults to the deployed Worker (set it
only to point at your own mirror), and X402_WALLET_PRIVATE_KEY is optional — set it
to auto-settle paid tools (HTTP 402) via x402; leave it empty to run only the
free/preview tools. The client reads both straight from process.env at startup.
// stdio client config (local path)
{
"mcpServers": {
"grey-ridge-x402-local": {
"command": "node",
"args": ["/absolute/path/to/mcp-client/dist/server.js"]
}
}
}
Call it (HTTP)
BASE=https://x402-data-api.sigrunner.workers.dev
# free preview — real sample, no payment
curl -s "$BASE/crypto/funding/preview"
# paid — returns HTTP 402 with an x402 payment-required challenge until you attach payment
curl -i "$BASE/crypto/funding?limit=10"
Any x402-capable client (e.g. the x402 libraries)
handles the 402 → sign → retry automatically.
Measured status
Uptime and latency per safety route are measured from a probe cron and published in docs/STATUS.md (regenerated hourly; window stated in the file, not an SLA).
How it runs
A single Cloudflare Worker implements every HTTP endpoint and the MCP server
behind a method-aware x402 gate. All upstreams are free/keyless public APIs (DefiLlama,
Hyperliquid, OKX, Polymarket Gamma, Base JSON-RPC, NVD/EPSS/CISA-KEV, crt.sh/RDAP/DoH); every
address/hash/domain input is validated and SSRF-guarded (fixed upstream hosts, no user-supplied
targets except the explicitly-scoped /scan/mcp and enrichment endpoints).
License
MIT — see LICENSE.
Agent-framework plugins (published)
Pre-trade token guard (/gate/swap, $0.008 USDC via x402 on Base) — drop-in, fails closed, pays inline per call (no API key, no account). Free sample: curl "https://x402-data-api.sigrunner.workers.dev/gate/swap/preview?token=degen"
| Framework | Install | Source |
|---|---|---|
| ElizaOS | npm i github:rezearcher/plugin-swap-gate | integrations/elizaos-plugin-swap-gate |
| LangChain.js | npm i github:rezearcher/langchain-swap-gate @langchain/core zod | integrations/langchain-swap-gate |
| Vercel AI SDK | npm i github:rezearcher/vercel-ai-swap-gate ai zod | integrations/vercel-ai-sdk-swap-gate |
Install from git — this always pulls the current build (keyless free preview + inline x402, fails closed only if both are unreachable) and compiles on install via prepare. Each also exposes a tokenReport tool wrapping /chain/token-report for standalone token screening. Minimal copy-paste examples live under docs/examples/{swap-gate,langchain,elizaos}/.
Market-data buyer (for agents that buy a price and must not present stale data as real-time — the Decision402 pattern): examples/market-data-decision-buyer.mjs. Runs the full agentic loop — reads the live 402 quote, enforces a per-call USDC budget, proves the feed via the free preview, pays /crypto/prices ($0.001 on Base), then rejects any row older than MAX_AGE_SECONDS. Steps 1-3 run with zero npm installs: node examples/market-data-decision-buyer.mjs bitcoin,ethereum.
Python / LangChain / CrewAI
pip install "git+https://github.com/rezearcher/greyridge-x402-tools" — token_report() and swap_gate() helpers (free preview by default, paid x402 mode optional) plus LangChain tools. Guides: https://rezearcher.github.io/greyridge-x402-tools/
Try it on YOUR token, free
curl "https://x402-data-api.sigrunner.workers.dev/chain/token-report/preview?token=0xYourBaseToken" returns the same full-depth report as the paid route ($0.01). /gate/swap/preview?token=0x… gives 3 free real-token GO/HOLD/BLOCK checks/day/IP.
Agent skill (drop-in SKILL.md for OpenClaw/Hermes-style agents): https://github.com/rezearcher/greyridge-token-safety-skill
CI gate for MCP servers
uses: rezearcher/greyridge-x402-tools@v0.3.3 with url: runs the free MCP tool-poisoning preview in CI and fails on a block verdict. Every run also emits a copy-paste README injection-security badge (badge-markdown output / step summary; commit-badge: true auto-commits it between ` markers, degrade-open) — the OWASP-MCP (LLM01/LLM08) dimension generic repo-quality badges omit, re-checked on every CI run and linked to the server's live security report. Listed on the [GitHub Marketplace](https://github.com/marketplace/actions/mcp-server-tool-poisoning-scan). Seeactions/mcp-scan/README.md`.
Python note: the default
Python-urllibUser-Agent is rejected with 403 at the Cloudflare edge. Userequests/httpx, or set any customUser-Agentheader.
Questions fréquentes
Qu'est-ce que ai.greyridgesignals/tech-risk ?
Free MCP security scan: tool-poisoning & prompt-injection audit (OWASP LLM01/08) + trust badge
Comment installer ai.greyridgesignals/tech-risk ?
Il fonctionne comme serveur distant. Dans Claude Code : claude mcp add --transport http tech-risk 'https://api.greyridgesignals.ai/mcp'. Pour les autres clients, collez la configuration JSON de cette page dans leurs réglages MCP.
Faut-il une clé API pour ai.greyridgesignals/tech-risk ?
Son entrée au registre n'indique aucune clé ni secret obligatoire. Certaines fonctions peuvent tout de même demander un compte auprès du service auquel il se connecte.
ai.greyridgesignals/tech-risk est-il open source ?
Oui. Son code est sur GitHub sous licence MIT.
Qui maintient ai.greyridgesignals/tech-risk ?
Il est publié par rezearcher. AI.tn le référence sans le maintenir ni le recommander.
Issu du registre MCP officiel. Être listé n'est pas une recommandation : vérifiez le code et les permissions d'un serveur avant de lui donner accès à vos données.