Email verify (powered by OpenMerch)

Verify whether an email address is deliverable, and surface basic reputation signals (disposable / free-provider / role address, a quality score), for a single address.

This skill uses OpenMerch to plan, route, and execute the check. You give it one email address; it returns a normalized result plus the raw provider output.

Requirements & cost

This skill uses OpenMerch to plan, route, and execute the email check. OpenMerch is an external paid API: you need an OpenMerch account and an API key, and each verification consumes your OpenMerch account balance/credits. The exact price is confirmed by the /v1/plan call before anything runs — that quote is the source of truth for what you pay (currently roughly $0.006–$0.007 per email, depending on OpenMerch pricing policy). The skill never charges more than the planned max_cost. ClawHub does not handle this billing and takes no fee — the charge is between you and OpenMerch.

Get a key from the Developer page in the OpenMerch app and set it as an environment variable:

export OPENMERCH_API_KEY="om_live_xxxxxxxx"
# Optional — defaults to https://api.openmerch.dev:
# export OPENMERCH_BASE_URL="https://api.openmerch.dev"

What this calls

No hidden network behavior. This skill makes only these OpenMerch HTTP calls, in order:

1. POST {OPENMERCH_BASE_URL}/v1/plan — confirm the job is executable and get the price.
2. POST {OPENMERCH_BASE_URL}/v1/execute — run the verification (one job).
3. GET {OPENMERCH_BASE_URL}/v1/jobs/{job_id} — only if the job is still executing, to poll until it finishes.

Every request sends the header X-OpenMerch-Key: $OPENMERCH_API_KEY. POST bodies send Content-Type: application/json. Base URL is OPENMERCH_BASE_URL or https://api.openmerch.dev.

How to run

You can run this two ways. Both do exactly the same calls.

Option A — reference script (deterministic)

Requires Node 18+ (uses built-in fetch; no npm install):

node verify-email.mjs [email protected]

It prints a JSON result to stdout (see "Output" below) and exits non-zero on error.

Option B — agent-driven (instructions)

If you (the agent) are running the flow with your own HTTP tools, follow these steps exactly:

1. Plan. POST /v1/plan:

{ "job_type": "email_reputation_v1", "input": { "email": "\x3CEMAIL>" } }

• If the response can_execute is not true, stop and report the reason. Do not execute.
• Set max_cost = quoted_customer_price_microcents if present, otherwise estimated_cost.max_microcents. /v1/plan is the source of truth for the price — never hardcode one.

2. Execute. Generate one UUID v4 as idempotency_key for this submission. POST /v1/execute:

{
  "job_type": "email_reputation_v1",
  "input": { "email": "\x3CEMAIL>" },
  "max_cost": \x3Cmax_cost from step 1>,
  "idempotency_key": "\x3Cuuid>"
}

• Reuse the same idempotency_key if you retry this same submission (e.g. after a timeout). Generate a new key only for a genuinely new verification. This prevents double charges.

3. Poll only if needed. If the execute response status is "executing", poll GET /v1/jobs/{job_id} every ~1s (cap ~8 tries / ~15s) until status is completed, failed, or cancelled. Most runs return completed immediately and no polling is needed.

4. Report. On completed, present the normalized result (below). On failed/cancelled, report error.code and error.message. Always report cost.total_microcents (the actual amount charged) and the job_id.

curl equivalent

BASE="${OPENMERCH_BASE_URL:-https://api.openmerch.dev}"

# 1. Plan
curl -sS -X POST "$BASE/v1/plan" \
  -H "X-OpenMerch-Key: $OPENMERCH_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"job_type":"email_reputation_v1","input":{"email":"[email protected]"}}'

# 2. Execute
# Replace \x3CMAX_COST_FROM_PLAN> with quoted_customer_price_microcents from /v1/plan,
# or estimated_cost.max_microcents if no quote was returned. Do not hardcode a price.
curl -sS -X POST "$BASE/v1/execute" \
  -H "X-OpenMerch-Key: $OPENMERCH_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"job_type":"email_reputation_v1","input":{"email":"[email protected]"},"max_cost":\x3CMAX_COST_FROM_PLAN>,"idempotency_key":"'"$(uuidgen)"'"}'

# 3. Poll only if status was "executing"
# curl -sS "$BASE/v1/jobs/\x3Cjob_id>" -H "X-OpenMerch-Key: $OPENMERCH_API_KEY"

Output

The skill returns a normalized object. The summary fields are best-effort and appear only when the underlying provider response contains them; raw is always the full, unmodified OpenMerch job output and is the source of truth. cost_usd is derived from the actual charged cost.total_microcents.

{
  "email": "[email protected]",
  "summary": {
    "deliverable": true,
    "is_disposable": false,
    "is_free_email": false,
    "quality_score": 0.95
  },
  "raw": { "...": "verbatim OpenMerch job output" },
  "cost_usd": 0.007,
  "job_id": "…"
}

Notes & limits

• One email per run. For a list, call the skill once per address.
• The skill executes a single atomic OpenMerch job — no multi-step orchestration.
• All monetary values from OpenMerch are in microcents (1 cent = 100,000 microcents; $1.00 = 10,000,000). cost_usd is cost.total_microcents / 10000000.