Skip to main content
Neural Draft uses a single-meter, pay-as-you-go model. Every operation costs credits. You either subscribe to a tier (which buys credits at a discount plus a monthly allowance), top up ad-hoc, or stay free. There’s no per-seat pricing, no per-tenant tax, no “Pro” gate. Every endpoint works at every tier; only the cost-per-credit differs.
Hard stop at zero credits. When your balance hits zero, every credit-consuming operation returns 402 insufficient_credits. Read-only endpoints continue to work. No overage. No invoice surprises.

​
What every operation costs

OperationCredits
Chat message1
Content update (PUT /v1/content/{key})1
Page create or update (POST/PATCH /v1/pages)1
Image register (URL swap or multipart upload)1
Page edit (inline AI tweak)2
Social post — text only5
Translate (per target language)7
Content plan creation10
Image generation32
Social post — text + image5 + 32
Social post — text + video5 + 40
Video clip — budget (Wan 2.1, default)40
Blog post (research + draft + featured image)60
Video clip — premium (Kling v2.1 / Runway Gen4)300
Website generation (full landing, multi-section)240
Newsletter subscribe / contact-form submit0
Quality-default models. Text generation runs on Claude Sonnet 4.6 (with Haiku 4.5 for high-volume conversational ops like chat and inline page edits); image generation runs on Google’s Nano Banana (Gemini 2.5 image). Premium video uses Kling v2.1 / Runway Gen4. Forms (newsletters & contact submissions) are free — they’re storage + delivery, not AI.
Premium video is gated. tier=premium on POST /v1/videos (Kling / Gen4) requires a Build plan or higher; Hobby and Free projects can use the budget tier (Wan 2.1) by default and will get 402 plan_required if they ask for premium.
Read-only endpoints (GET) and a handful of lightweight writes (e.g. updating a draft, attaching a tag, scheduling a post, registering a component) do not consume credits. Mutating writes through the public API — PUT /v1/content/{key}, POST/PATCH /v1/pages, and image_register (URL swap on PUT /v1/images/{key} or multipart upload on POST /v1/images) — each cost 1 credit, charged on the successful write.
Social post media surcharge. POST /v1/social-posts with type=ai always charges the 5-credit text base; if media_type=image or media_type=video is also set, an extra image (32) or video (40, budget tier) charge is recorded against the same job, so the total reflects what was actually generated. Social posts always use the budget video tier — request premium clips via POST /v1/videos with tier=premium (Build+).
A few examples to ground the numbers:
  • A blog post with 4 translations: 60 + 4 × 7 = 88 credits.
  • A cross-platform social post with one image (single API call, posts to FB + IG + TikTok): 5 + 32 = 37 credits.
  • A landing page with a hero image generated mid-session: 32 + 240 = 272 credits.
  • A short product reel via POST /v1/videos (budget, default) or media_type=video social: 40 credits per clip. Premium clip via tier=premium: 300 credits.

​
Plans

A plan doesn’t unlock features — every operation is available at every tier. A plan buys credits at a discount with a monthly allowance.
PlanCredits / moPrice$/creditDiscount
Free250$0——
Hobby5,000$25$0.005base
Build25,000$119$0.004765% off
Scale100,000$399$0.0039920% off
Enterprisecustomcustomvolume contractup to 40% off

​
Top-ups

Top up anytime, on any tier (including Free), at the base rate of $0.005 / credit. No commitment, no expiry while the project is active.
BundleCredits
$204,000
$10020,000
$500100,000

​
Free tier

Every project starts with 250 credits per month, no card required. The free tier is sized for evaluation — enough to build, test and demo a working integration end to end, not enough to run a real workload. Mix-and-match examples of what 250 credits buys at the new quality-default pricing:
  • ~250 chat messages, or
  • ~50 text-only social posts, or
  • ~7 Nano Banana images, or
  • 4 full Sonnet 4.6 blog posts, or
  • ~6 budget video clips, or
  • 1 full website generation.
Premium video (tier=premium) is not available on Free or Hobby — it requires a Build plan or higher. Free-tier projects also surface a small “Built with Neural Draft” badge in embedded widgets and on platform-rendered pages — turning that off requires a paid tier. Hit zero and the API returns 402 insufficient_credits until next month — no surprise bills. Top up with a one-time pack or upgrade to a paid tier whenever you’re ready.

​
Tracking usage

The dashboard shows a real-time credit ledger. The same data is on the API: 
curl https://api.neuraldraft.io/v1/projects/me/usage \
  -H "Authorization: Bearer $NEURALDRAFT_API_KEY"

{
  "period_start": "2026-04-01T00:00:00Z",
  "period_end": "2026-05-01T00:00:00Z",
  "plan": "build",
  "credits_included": 25000,
  "credits_used": 8421,
  "credits_remaining": 16579,
  "top_ups": 0,
  "by_operation": {
    "blog_post.generate": 5200,
    "image.generate": 1840,
    "social_post.generate": 1100,
    "translation.batch": 281
  }
}
A few patterns we’ve seen work well:
  • Soft alerts: poll GET /v1/projects/me/usage and Slack the team when credits_remaining drops below 20% of the included allowance.
  • Per-customer cost tracking: tag your own end-customers in metadata.customer_id on async jobs; the dashboard exports a CSV grouped by metadata.
  • Hard ceilings per environment: production keys have full scopes; preview keys have content:* only and live on a separate top-up budget.

​
How pricing changes

We re-evaluate credit costs roughly every 90 days against current upstream model prices (OpenAI, Anthropic, image and video providers). When upstream list prices drop materially, your credit costs drop too. When they rise, we may need to bump credit costs — always with at least 30 days’ notice before any change takes effect, communicated by email and on the changelog. Predictable, in plain words:
  • Per-credit pricing tracks upstream list-price moves and major capability shifts (e.g. a new model tier). Nothing else.
  • Increases are pre-announced with 30+ days’ notice. You always have time to budget or move.

​
FAQ

Async jobs (blog generation, batch translation, etc.) reserve credits when queued. If your balance is too low at queue time, the request returns 402 insufficient_credits immediately — nothing is started. Reserved credits are refunded if the job fails for our reasons (e.g. upstream timeout).
Unused subscription credits don’t roll over or refund — that’s how the discount is funded. Top-up credits are refundable for 30 days at our discretion if your project is in good standing and you haven’t used >25% of the bundle. Email info@neuraldraft.io.
Failed AI generations that we retry server-side don’t cost extra — you pay once per successful or definitively-failed job. Webhook deliveries that retry due to your endpoint returning non-2xx are billed per delivery attempt (0.02 credits each); we cap retries at 5 with exponential backoff.
No. GET endpoints don’t consume credits unless explicitly noted in the API reference for that endpoint (very few do). Rate limits still apply — see rate-limits.
Agencies managing multiple projects get a master billing account — every project rolls up into a single invoice with project-level breakdowns. Email info@neuraldraft.io for volume terms.
Subscriptions: monthly, on the day-of-month you signed up. Top-ups: on purchase. Storage and recurring line items (connected accounts, custom domains): metered hourly, billed at month-end against the included allowance, then top-ups, then card-on-file (only if you’ve opted into auto-recharge).