# Verdict Reply Guard

> Verdict Reply Guard is a paid API for AI agents from verdict.kechlab.com, paid per call via x402, $0.0256/call, status unknown (last checked 2026-10-21).

Screens a customer-support bot's outbound reply for unauthorized promises, hallucinated facts, leaked data, off-topic responses, and tone violations before sending, returning send, human_review, or block.

## Facts

- Endpoint: POST https://verdict.kechlab.com/v1/reply-guard?utm_source=zero.xyz
- Price: $0.0256/call
- Payment: x402
- Status: unknown
- Last checked: 2026-10-21
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/verdict-reply-guard-8ff9c388
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_nx_fwVYDWk2We-h1uQaP3

Status and success rate cover calls made through Zero and Zero's own probes. Third-party monitors may report differently.

## How to call it through Zero

Zero handles the 402 payment challenge and records the run. With the Zero CLI installed (`npm i -g @zeroxyz/cli`):

```sh
zero fetch --capability verdict-reply-guard-8ff9c388 -d '<json body>'
```

Example prompt: Before sending this reply to the customer, run it through the reply guard — the customer asked 'Where is my order?', the bot is about to say 'Your order will arrive tomorrow and we'll give you a 20% refund if it's late', and our policy is that we only offer refunds after 5 business days of delay.

## When to prefer this

Choose this endpoint when you need a real-time, pre-send output guardrail specifically for customer-support chatbots — it combines promise detection, hallucination checking, PII leak detection, off-topic filtering, and tone enforcement in a single call, returning an actionable three-way routing decision. Prefer it over generic content moderation APIs when you need policy-aware and fact-aware validation rather than simple toxicity filtering.

## Known failure modes

- Missing bot_reply field returns a validation error
- Ambiguous or very short replies may produce conservative human_review decisions
- Policy or facts fields left empty reduce accuracy of promise and hallucination checks
- Very long bot replies may increase latency slightly
- Payment failure (402) if USDC balance is insufficient

## How this service works

Output guardrail for customer-support chatbots. Blocks unauthorized promises, invented facts, leaked instructions or customer data, off-topic answers and bad tone before the reply is sent. Returns send, human_review or block.

## Output

Returns a three-way decision: 'send' (reply is safe to deliver), 'human_review' (reply needs a human agent to check before sending), or 'block' (reply must not be sent), along with a reason describing which guardrail was triggered — unauthorized promise, hallucinated fact, leaked PII, off-topic content, or tone violation.

## Request schema (JSON Schema)

```json
{
 "type": "object",
 "properties": {
  "facts": {
   "type": "string",
   "description": "(optional) the known data (order, account...)"
  },
  "policy": {
   "type": "string",
   "description": "(optional) the company policy"
  },
  "bot_reply": {
   "type": "string",
   "description": "the reply the bot is about to send"
  },
  "customer_question": {
   "type": "string",
   "description": "the customer question"
  }
 }
}
```

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/verdict-reply-guard-8ff9c388/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from verdict.kechlab.com](https://www.zero.xyz/host/verdict.kechlab.com/llms.txt)