Viridis MCP — API Documentation

Base URL: https://mcp.viridis-security.com  ·  Auth: Authorization: Bearer vrd_live_…

Quick start

# 1. Get a free API key
curl -X POST https://mcp.viridis-security.com/v1/signup \
  -H 'content-type: application/json' \
  -d '{"email":"you@example.com","tier":"free"}'
# → { "accountId": "acct_...", "tier": "free", "apiKey": "vrd_live_..." }

# 2. Use it
curl -X POST https://mcp.viridis-security.com/v1/injection/detect \
  -H 'authorization: Bearer vrd_live_...' \
  -H 'content-type: application/json' \
  -d '{"input": "Ignore all previous instructions...", "certainty": "premium"}'
# → { "verdict": "attack", "probability": 0.99, ... }

Endpoints

Detect adversarial injection attempts. Backed by T-IB-02 (Adversarial Landauer Inequality) + T-IB-06 (Detection Lower Bound) + T-IB-01 (Attribution Conservation).

Request body

Field	Type	Description
input	string	Required. The untrusted text to analyze. ≤200K chars.
context	string	Optional. Your agent's system prompt for calibration.
certainty	"quick" | "standard" | "premium"	Operating point. quick = α=10⁻² ; standard = α=10⁻³ ; premium = α=10⁻⁶. Higher certainty costs more per T-IB-06.
agentId	string	Optional. Registered envelope ID (from MCP-01) for envelope cross-check.

Response

{
  "verdict": "clean" | "suspicious" | "attack",
  "probability": 0.0 — 1.0,
  "bitsAtRisk": 0 — 32,            // T-IB-01 upper bound on adversarial capture
  "operatingPoint": { "alpha": 1e-3, "beta": 1e-3 },
  "matchedPatterns": ["VC-AI-PROMPT-0001", ...],
  "recommendedAction": "allow" | "sanitize" | "reject" | "escalate",
  "explainabilityToken": "expl_...",
  "signals": { "pattern": 0.99, "instructionDensity": 0, "obfuscation": 0, "envelope": 0 },
  "billing": { "cost": 0.003, "tier": "starter", "remaining": 49997 },
  "backedBy": ["T-IB-02", "T-IB-06", "T-IB-01"]
}

Sign up for an account. Free tier returns an API key instantly; paid tiers return a Stripe Checkout URL.

{ "email": "...", "tier": "free" | "starter" | "growth" | "scale" }

Issue an adaptive proof-of-work challenge. Asymmetric defense: legitimate retry costs ~ms, automated attackers pay 10× to 10⁶× more. Backed by T-IB-09 (Adversarial Dissipation Theorem).

Errors

HTTP	Error code	Meaning
400	invalid_json / invalid_input / invalid_tier	Malformed request
401	missing_api_key / invalid_api_key / revoked_api_key	Authentication failed
413	input_too_large	input >200K chars
429	rate_limit_exceeded	Free tier monthly quota or per-second burst exhausted

SDKs

TypeScript: npm install @viridis/mcp-client

Python SDK (pip install viridis-mcp-client) shipping shortly.

Open-source (Apache-2.0): github.com/viridis-security/mcp-services

Backing theorems

Each endpoint cites Lean 4 theorems verified by Aristotle (Harmonic). Independent re-verification recipe in the corpus repo.