The Molixa API platform: two AI APIs, one prepaid balance
Molixa is a B2B API platform with two production APIs behind a single prepaid credit balance. Buy credits once, spend them across watermark removal and AI content generation, and pay nothing when your calls fail.
No subscription. Prepaid credits. One balance for every API.
# One key. One balance. Two APIs.
curl https://molixa.app/api/v1/content/generate \
-H "X-Api-Key: mlx_live_…" \
-d '{"task":"product-description"}'
curl https://molixa.app/api/v1/images/watermark-remove \
-H "X-Api-Key: mlx_live_…" \
-F "[email protected]"What the Molixa API platform is
Molixa is a B2B API platform for teams that need AI image and text features in their own product without building or operating the models. Instead of standing up inference infrastructure, wiring together vendors, and reconciling three separate invoices, you get two ready REST APIs behind one account, one balance, and one bill.
There are two products today. The Watermark Remover API takes an image and returns a clean one, with the watermark or unwanted object erased by auto-detection plus AI inpainting. The AI Content Writer API takes a prompt and returns finished copy: blog sections, product descriptions, ad text, emails. Both are plain HTTPS endpoints. There is no SDK to install and no framework to adopt. If your language can make an HTTP POST, it can call Molixa.
The point of the platform is consolidation. Your credit balance, API keys, usage, billing, rate limits, and audit trail live in one place, the partner dashboard at /dashboard/platform. Enabling a second product does not mean a second contract, a second minimum, or a second top-up flow. It mints a key and starts drawing from the same balance you already funded.
- Two REST APIs: watermark removal and content generation
- One prepaid credit balance funds both, with one invoice
- No SDK, no framework lock-in, server-to-server over HTTPS
- Scoped API keys, per-partner rate limits, and audit logs built in
- Prepaid credits, no subscription and no per-seat fees
# one key · two APIs · one bill
# 1 — clean an image
POST /api/v1/images/watermark-remove
# → clean CDN url, ~2s, 80 cr
# 2 — write some copy
POST /api/v1/content/generate
# → finished text, 0.20–0.50 cr/wordThe two APIs at a glance
Both APIs share the same authentication, the same billing model, and the same error format, so once you have integrated one, the second is a short step. The difference is what they do and how they charge.
The Watermark Remover API is a single call: POST an image (PNG, JPEG, or WebP up to 12MB, or a JSON imageUrl) to /api/v1/images/watermark-remove. You can pass exact boxes for a known watermark region, a white-on-black mask, or nothing at all to let auto-detection find it. It returns a CDN URL for the cleaned image in about two seconds and charges a flat 80 credits ($0.08) per finished removal. Auto-detection is conservative by design: if it cannot find a clear watermark it returns NOT_DETECTED, keeps your original untouched, and refunds the credit. Full reference lives at /watermark-api, and the product page is at /platform/watermark-remover-api.
The AI Content Writer API generates copy from a prompt at /api/v1/content/generate. You choose a task (blog intro, product description, ad copy, email, freeform, and more), a tone, a language, and a target length up to 2,000 words. Two tiers control the model and the price: the fast tier runs on DeepSeek at 0.20 credits per word ($0.20 per 1,000 words), and the quality tier runs on OpenAI at 0.50 credits per word ($0.50 per 1,000 words). Full reference lives at /content-api, and the product page is at /platform/ai-content-writer-api.
- Watermark Remover: POST an image, get a cleaned CDN URL back, 80 credits per removal ($0.08)
- Content Writer: POST a prompt, get finished copy, 0.20 credits/word fast ($0.20/1k) or 0.50 quality ($0.50/1k)
- Same auth (X-Api-Key), same JSON error codes, same prepaid balance
- Docs: /watermark-api and /content-api; product overviews under /platform
Refunded on failure or unused words
One credit balance, both APIs, prepaid
The unified credit model is the core idea. A credit is a unit of value fixed at $0.001. You buy a pack of credits up front, and every call to either API debits that one shared balance. There is no separate wallet for images and text, and no monthly subscription to keep active. When the balance runs low, you top up; when it hits zero, the API returns INSUFFICIENT_CREDITS until you do.
Because both products draw from the same pool, credits flow to wherever your workload actually is. A marketplace that mostly cleans listing photos and occasionally drafts descriptions does not need to guess a split in advance. Buy credits, and spend 80 on a removal here or a fraction of a credit per word there, from the same number.
Billing is metered honestly. The Watermark API charges only on a successful, returned removal; a NOT_DETECTED or a processing failure is refunded. The Content API authorizes the worst case (your maxWords) up front, generates, then refunds the unused remainder, so you are billed for the words you actually receive and never more. A blocked or failed generation is fully refunded. You are never charged for output you did not get.
- 1 credit = $0.001, one balance shared by both APIs
- Prepaid packs, no subscription, no per-seat pricing
- Watermark charges only on a successful removal; auto NOT_DETECTED is refunded
- Content authorizes maxWords, then refunds the unused words after generation
- Failed or policy-blocked calls are refunded in full
How it works end to end
Getting from zero to a live integration is a short, linear path. You apply, you get approved, you enable a product, you get a scoped key, you call the API, and you top up when you need more. Here is each step.
- 1
Apply at /partners
Onboarding is by application; each partner is reviewed by hand, usually within one to two business days.
/partners - 2
Get approved
Once approved, open the partner dashboard where your balance, keys, and usage live.
/dashboard/platform - 3
Enable a product
Enabling an API mints a key scoped to that product; a wrong-product call is rejected.
SCOPE_FORBIDDEN - 4
Call the API
Send your key in the header — secret mlx_ server-side, publishable pub_ domain-locked in the browser.
X-Api-Key - 5
Top up when low
Buy another pack from the dashboard and keep going — no plan change, no re-integration.
credits
Two key types: secret and publishable
Molixa gives you two kinds of key so you can call the platform safely from either side of your stack. Both go in the same X-Api-Key header; the prefix tells the platform how to treat them.
The secret key starts with mlx_ and is for server-to-server calls. It is never stored in cleartext: the platform keeps only a SHA-256 hash and looks up your key by that hash, so even a database read cannot recover the raw key. Keep it on your backend, out of client code and out of your repository. Lost or exposed it? Rotate a fresh one from the dashboard at any time, and the old one stops working immediately.
The publishable key starts with pub_ and is safe to put in a browser. It is domain-locked: the platform checks the request Origin against the list of domains you authorized for that key, and rejects anything else with ORIGIN_NOT_ALLOWED. That lets an embedded widget call the API directly from your users' browsers without ever shipping your secret key to the client. Use the secret key for backend work and the publishable key only where a browser needs to call in.
- mlx_ secret key: server-side only, stored as a SHA-256 hash, rotatable any time
- pub_ publishable key: safe in the browser, locked to your authorized domains
- Wrong product for a key returns SCOPE_FORBIDDEN; wrong domain returns ORIGIN_NOT_ALLOWED
- Both authenticate through the same X-Api-Key header
Why build on Molixa
The honest case for Molixa is cost, consolidation, and control, not a claim to have the only model that works. On watermark removal, a flat 80 credits ($0.08) per finished image sits under remove.bg's per-image floor, and you only pay when a clean image comes back. On content, calling Molixa's fast tier at $0.20 per 1,000 words means you are not paying OpenAI list rates directly and not maintaining the retry, moderation, and billing plumbing around them yourself.
Consolidation is the practical win. Two capabilities, one integration pattern, one credit balance, one invoice, one dashboard. A team that would otherwise sign up for an image API and a text API separately, learn two auth schemes, and reconcile two bills, instead learns one and reconciles one. Enabling the second product is a dashboard toggle, not a new procurement cycle.
There is no subscription lock-in. Credits are prepaid and one-time; you are not on a monthly plan that bills whether you use it or not, and there are no per-seat fees. And there is a real privacy and compliance posture behind it. Both APIs require an ownership or acceptable-use affirmation per call, content runs through moderation, and there is an Acceptable Use Policy that binds how the tools may be used. You are building on infrastructure that is designed to say no to abuse, which protects your account as much as ours.
Who it is for
The platform fits any product that handles user images or generates text at scale and would rather integrate than build. A few shapes come up repeatedly.
Marketplaces and classifieds use the Watermark Remover so sellers can clean stock or competitor watermarks off photos they own the rights to, and the Content Writer to auto-draft listing descriptions from a few fields. The detect-gated, non-destructive flow (show a Remove watermark button, process on click, always keep the original) fits a listing pipeline cleanly.
SaaS products embed either API as a feature inside their own app: a design tool that cleans up uploaded assets, a CRM that drafts outreach emails, a help desk that summarizes tickets. Agencies and e-commerce brands run both at volume, cleaning product photography and generating product copy from one balance, and bill the credits back against client work. Because pricing is prepaid and usage-metered, a small pilot and a large production workload use the exact same integration; only the pack size changes.
Marketplaces
Clean listing photos sellers own the rights to, and auto-draft descriptions from a few fields.
SaaS products
Embed image cleanup or copy generation as a native feature inside your own app.
Agencies
Run both APIs at volume and bill the credits back against client work.
E-commerce
Product photography and product copy generated from one prepaid balance.
Security and reliability
The platform is built so that a mistake, a leaked key, or a burst of traffic does not turn into a loss or an outage for you. Secret keys are never stored in cleartext; the platform keeps a SHA-256 hash and matches by hash, so the raw key exists only on your side. Keys are scoped per product and, for publishable keys, locked to your domains, so the blast radius of any single key is bounded.
Billing is atomic and can never go negative. Every debit is applied in a way that cannot overshoot your balance or race itself into a negative number, which is what makes the refund-on-failure behavior safe: authorizing, charging, and refunding are consistent even under concurrent calls. The Content API accepts an Idempotency-Key header so a network retry replays the first result instead of generating and charging twice.
Every partner gets per-partner rate limits, so one noisy integration cannot starve another, and a burst is answered with a 429 RATE_LIMITED and a short back-off rather than dropped work. Content prompts and outputs pass through moderation; flagged text is blocked and not charged. Actions are recorded in audit logs so you have a trail of what was called and billed. And every error is a JSON 4xx response with a stable code (INSUFFICIENT_CREDITS, SCOPE_FORBIDDEN, NOT_DETECTED, BLOCKED, RATE_LIMITED, and the rest), so your integration can branch on a string that will not change out from under it.
Transparent pricing and credit packs
There is one price list, and it is the same one the billing code uses. A credit is worth $0.001. A watermark removal costs 80 credits, which is $0.08. Content costs 0.20 credits per word on the fast tier ($0.20 per 1,000 words) and 0.50 credits per word on the quality tier ($0.50 per 1,000 words). Nothing is metered or rounded in a way you cannot see: you can compute your own cost before you send a call.
Credits are sold in one-time packs, and bigger packs cost less per credit. The Starter pack is $50 for 50,000 credits ($0.00100 per credit). The Growth pack is $90 for 99,000 credits, a 10% bonus that works out to about $0.00091 per credit. The Scale pack is $230 for 276,000 credits, a 20% bonus at about $0.00083 per credit. Running serious volume beyond that? Enterprise pricing is custom; mention your expected monthly usage when you apply and we will tailor a pack.
| Unit | Credits | USD |
|---|---|---|
| 1 credit | 1 | $0.001 |
| Watermark removal | 80 | $0.08 |
| Content · fast | 0.2 / word | $0.20 / 1k |
| Content · quality | 0.5 / word | $0.50 / 1k |
| Starter · 50,000 credits | 625 removals · or 250,000 fast-tier words · or any mix | |
- 1 credit = $0.001; watermark = 80 credits ($0.08); content = 0.20 cr/word fast, 0.50 quality
- Starter: $50 for 50,000 credits ($0.00100/credit)
- Growth: $90 for 99,000 credits, 10% bonus (about $0.00091/credit)
- Scale: $230 for 276,000 credits, 20% bonus (about $0.00083/credit)
- Enterprise pricing is custom; one-time packs, bigger packs cost less per credit
One balance, honest pricing
Prepaid credit packs power both APIs. One credit is $0.001, bigger packs cost less per credit, and you only pay for calls that succeed.
- ~625 watermark removals
- ~250,000 words of content
- $0.0010 per credit
- ~1,237 watermark removals
- ~495,000 words of content
- $0.0009 per credit
- ~3,450 watermark removals
- ~1,380,000 words of content
- $0.0008 per credit
Need enterprise volume? Talk to us.
Getting started
The fastest way in is to read the reference for the API you care about and then apply. The Watermark Remover API reference is at /watermark-api and the AI Content Writer API reference is at /content-api. Both show the endpoint, the request and response shapes, the error codes, and copy-paste cURL and Node examples.
- 1
Read the docs
Endpoint, request and response shapes, error codes, and cURL + Node examples.
/watermark-api · /content-api - 2
Apply for access
Approval is usually one to two business days.
/partners - 3
Enable + first call
Enable a product, mint a scoped key, and make your first call the same afternoon.
mlx_ - 4
Manage everything
Keys, credits, usage, and top-ups all live in the dashboard.
/dashboard/platform
Frequently asked questions
It is a B2B API platform with two production REST APIs, watermark removal and AI content generation, behind a single prepaid credit balance. You apply for access, enable the products you want, get scoped API keys, and call plain HTTPS endpoints. One account, one balance, one bill for both tools.
Yes. There is one prepaid balance, and both APIs debit it. A credit is worth $0.001. A watermark removal costs 80 credits ($0.08); content costs 0.20 credits per word on the fast tier and 0.50 on the quality tier. Buy a pack once and spend it across images and text in any mix you like.
No. Credits are prepaid and sold in one-time packs with no subscription and no per-seat fees. You top up when the balance runs low, and unused credits stay yours. When the balance hits zero the API returns INSUFFICIENT_CREDITS until you buy more.
You authenticate with the X-Api-Key header. A secret mlx_ key is for server-side calls and is stored only as a SHA-256 hash, never in cleartext. A publishable pub_ key is safe in the browser and is locked to the domains you authorize. Keys are also scoped per product: a key issued for one API cannot call the other, and a wrong-product call returns SCOPE_FORBIDDEN.
A watermark removal is a flat $0.08 and is billed only when a clean image is returned, which sits under remove.bg's per-image floor. For content, the fast tier is $0.20 per 1,000 words, so you are not paying OpenAI list rates directly and not building the retry, moderation, and billing plumbing around them yourself. You also get both capabilities under one integration and one bill.
You are not charged for output you did not receive. A watermark auto-detection that finds nothing (NOT_DETECTED) and a processing failure are refunded, and your original image is never damaged. The Content API authorizes the target word count, then refunds the unused words; a failed or policy-blocked generation is refunded in full.
Apply at /partners. Onboarding is by application because it is a B2B program, and each partner is reviewed by hand, usually within one to two business days. Once approved, enable a product from the dashboard at /dashboard/platform to mint a scoped key and make your first call.
The Watermark Remover API reference is at /watermark-api and the AI Content Writer API reference is at /content-api. Each covers authentication, the endpoint, request and response formats, the stable error codes, and copy-paste cURL and Node examples.
Secret keys are stored as SHA-256 hashes, keys are scoped per product and domain-locked for browser use, and billing is atomic and can never go negative. Content requests support idempotency keys so retries do not double charge, every partner has per-partner rate limits, content passes through moderation, actions are audit-logged, and all errors are JSON 4xx responses with stable codes.
Ship AI features this week
Create an account, apply for API access, and start calling both APIs from a single prepaid balance.