Docs

cat > ad.json <<'EOF'
{
  "ad_copy": {
    "primary_text": "Still paying an agency $3k/mo for ads a robot writes better?",
    "headline": "Fire your ad agency"
  },
  "platform": "meta",
  "format": "static",
  "product_context": "AI Meta-ads automation tool, $99/mo",
  "target_audience": "DTC founders spending $2-20k/mo on Meta ads"
}
EOF

curl -sS -X POST https://www.spendict.com/api/v1/assess \
  -H "Authorization: Bearer spd_live_YOUR_KEY" \
  -H "Content-Type: application/json" \
  --data @ad.json

The body goes in a file on purpose: real ad copy is full of characters shells mangle (apostrophes, !, $99). Inside <<'EOF' nothing is interpreted, and --data @ad.json sends it byte-for-byte.

Input

The verdict

Scores are 0–100 across seven dimensions (hook, angle, clarity, audience_resonance, platform_fit, cta, compliance_safety) plus an authoritative, deterministic launch_recommendation:

• run — worth spending on: overall ≥ 78, hook ≥ 60, no hard gate tripped.
• fix_first — a real spine but a specific fixable problem; the fixes are the path to run. A dead hook or a likely policy rejection always caps an ad here.
• kill — structurally broken (two or more failing fundamentals); faster to regenerate than to fix.

{
  "overall_score": 78,
  "launch_recommendation": "run",
  "dimension_scores": {
    "hook": 78, "angle": 80, "clarity": 82,
    "audience_resonance": 80, "platform_fit": 85,
    "cta": 0, "compliance_safety": 95
  },
  "predicted_failure_mode": "No material weakness — would run as-is.",
  "prioritized_fixes": [],
  "enhancements": [
    { "issue": "No conversion path (CTA scored 0)",
      "suggestion": "Add one clear next step — optional; ships as-is." }
  ]
}

curl -sS -X POST https://www.spendict.com/api/v1/assess-batch \
  -H "Authorization: Bearer spd_live_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "platform": "meta",
    "product_context": "AI Meta-ads automation tool, $99/mo",
    "target_audience": "DTC founders spending $2-20k/mo on Meta ads",
    "variants": [
      { "label": "v1", "ad_copy": { "primary_text": "Still paying $3k/mo for ads a robot writes better?" } },
      { "label": "v2", "ad_copy": { "primary_text": "Introducing our new ad platform for teams." } }
    ]
  }'

Input

The response

Variants ranked by verdict (run > fix_first > kill) then overall score, a summary count, and per-variant errors (e.g. if quota runs out partway). Take the top N run and launch them.

{
  "summary": { "total": 2, "assessed": 2, "failed": 0,
    "run": 0, "fix_first": 1, "kill": 1, "top_label": "v1" },
  "ranked": [
    { "rank": 1, "label": "v1", "launch_recommendation": "fix_first", "overall_score": 74, ... },
    { "rank": 2, "label": "v2", "launch_recommendation": "kill", "overall_score": 46, ... }
  ],
  "errors": []
}