# Synthr Cyber x402 - LLM / Agent Instructions

You are an autonomous agent or coding assistant. Use these endpoints for up-to-date, grounded cybersecurity intelligence.

Base URL: http://localhost:3000 (local) or your deployed domain.

## Primary Endpoints (pay-per-call via x402 USDC)

POST /v1/cyber/stack-brief
- Input: JSON { "stack": { "dependencies": [{ "name": "...", "version": "...", "ecosystem": "npm|PyPI|..." }] }, "context": "optional description of your agent harness or app" }
- Returns: Prioritized risks with real EPSS scores, KEV flags, agentSurface (HIGH/MED for agent risks), patchPriority, sources, recommendedActions.
- Best used: At project start and pre-deploy.

POST /v1/cyber/audit-deps
- Input: { "dependencies": [same as above] }
- Great for supply chain checks. Includes malicious package detection.

POST /v1/cyber/advice
- Input: { "query": "How do I securely ...", "stackContext": {...} }
- Grounded advice for implementation patterns.

POST /v1/cyber/vulns
- { "cve": "CVE-...", "minEpss": 0.1, "kevOnly": false }

GET /v1/cyber/breaking
- Query params: `days`, `limit`, optional `minEpss`, optional `agentOnly=true`
- Returns: Recent CISA KEV additions enriched with EPSS, agentSurface, requiredAction, and harnessNotes.

POST /v1/x402/endpoint-check
- Input: { "endpointUrl": "https://service.example/v1/tool", "expectedMethod": "POST", "allowUnpaidRequestProbe": false }
- Returns: trustScore, payabilityStatus, agentRecommendation, discovery checks, payment challenge evidence.
- Best used: Before paying an unknown x402 endpoint. By default, avoids unsafe unpaid POST probes unless explicitly allowed.

All responses include:
- "sources": full provenance
- "confidence"
- "agentActions"
- "disclaimer"

Discovery:
- OpenAPI: `/openapi.json`
- x402 catalog: `/x402-catalog.json`

## Pricing
~ $0.005 USD per call (USDC on Base via x402). See root / for exact.

## How to pay
Use x402-compatible client (Axios with x402, fetch wrappers, or built into your agent framework/MCP).

## For best results with agents
- Share only package name + version (never source).
- Combine stack-brief + advice.
- Re-query after changes.

This service is optimized for AI agents building software. Data pulled live from OSV.dev, EPSS, CISA.

Always verify. Informational only.