Bulk image API · native MCP for AI agents

Render the images your whole campaign needs, from one template.

PixelDrive is a render API, not an AI art generator. Design a template once, then turn it into thousands of pixel-exact, on-brand variants from a single call. Sub-second per render, repeats cached free.

Start free Try the live demo Explore the API

pixeldrive.pro/dashboard

Bulk render

campaign_q3 · 5 batches

New batch +

TemplateVariantsStatusP50

diet-menu-ad4,800● done0.7s

podcast-cover320● done0.6s

listing-card12,400● running0.8s

cert-badge1,200● done5ms

hero-banner960○ queued—

Success rate▲

100%

The range, scrolling. Ads, listings, covers, posts — one API.one endpoint · sub-second each

Milestone
Brand quote
Ad creative
Webinar
Sale
Podcast cover
Metric
Brand quote
Hiring
Milestone
Brand quote
Ad creative
Webinar
Sale
Podcast cover
Metric
Brand quote
Hiring

Promo
Listing
Article
Launch
Sale
Newsletter
Pricing
Event
Quote
Promo
Listing
Article
Launch
Sale
Newsletter
Pricing
Event
Quote

Why it’s different

It does exactly what you tell it.

Most “image generation” means an AI guessing pixels from a prompt. PixelDrive is the opposite: your design, your data, rendered the same way every time. Here is where the alternatives break down.

Instead of AI art generators

Midjourney and DALL·E invent an image from a prompt. The result is a surprise every time — fonts drift, text turns to gibberish, your logo becomes a smudge.

PixelDrive renders your actual design. The price reads $49 because you set $49. Same input, same pixels, every single run.

Instead of Canva or Photoshop

Design tools are built for one person making one graphic. Eighty variants means eighty afternoons of duplicating layers and swapping text by hand.

Mark the parts that change as variables once. Then a single API call — or a CSV — produces all eighty, or ten thousand, with no human in the loop.

Instead of Bannerbear or Placid

The dynamic-image APIs had the right idea, but they meter per template, throttle throughput, and were never built for AI agents.

Sub-second renders on a fleet built for volume, previews and cache hits free, credits that roll over, and a native MCP server so agents generate real branded images — not hallucinations.

Deterministic
Sub-second p50
Cache hits free
Credits roll over
REST + MCP
1,000 free renders

Try it live

Type something. Watch it render.

A real render from the live API — the same endpoint you’d call. No sign-up to look.

Languagetranslated on render, free

Like it? Get 1,000 free renders →

Our main edge · Model Context Protocol

The image API your AI
can actually use.

PixelDrive is the only template-to-image API with a native MCP server. Drop one URL and a token into Claude, Cursor, or any MCP-aware client, and your agent designs, fills, previews and generates on-brand images from the chat. No SDK, no glue code, no hallucinated graphics.

01
It discovers your editor.
One call returns every element type, filter, font and figure the editor supports — so the model builds with what actually exists instead of guessing.
02
It builds and previews — free.
Compose a whole design in one call, align and stack without pixel math, then render a preview it can actually see. Previews never spend a credit.
03
It generates only when you mean it.
A single explicit action produces the final, billed image. Everything upstream — search, build, edit, preview — is free to iterate on.

~/.config/mcp.jsonconnected

{
  "mcpServers": {
    "pixeldrive": {
      "url": "https://pixeldrive.pro/mcp",
      "headers": {
        "Authorization": "Bearer pdrv_…"
      }
    }
  }
}

token → /dashboard/tokens

The toolset26 tools

Discover

describe_capabilitieslist_fontssearch_photossearch_icons

Build

create_templateadd_elementsalign_elementsupdate_element

Render

preview_image · freegenerate_image · 1 credit

Build, edit, search and preview are all free — the only billable action is the final generate_image.

The pipeline, briefly

Four steps from design
to ten-thousand-an-hour.

Every stage exists because we needed it at our own throughput. Nothing here is hypothetical.

01
Design
A template, made once.
Stage the design in the editor. Click the database icon on any element to expose it as a variable. The template is the only file that needs version control; everything downstream operates on it.
editor · starter.template3 variables marked
T
Aa
◯
▭
⬚
◐
⊞
Title — text variableYour title hereSubtitle — text variableSubtitle, also a variable
var · 3 fields
Variables
titletext
subtitletext
imagesrc
02
Call
A payload, posted in.
One template id plus the variables you want filled. Authenticated by Bearer token, billed per render, cached aggressively. There is no separate batch endpoint to learn — the API is one shape.
POST /v1/render200 OK · 1024ms
```
Authorization: Bearer pdrv_…
Content-Type:  application/json

{
  "templateId": "tpl_3b9b184a",
  "payload": {
    "title":    { "text": "10 EASY DIET MENU IDEAS" },
    "subtitle": { "text": "THE COOKING BEST" },
    "image":    { "src":  "https://cdn/photo.jpg" }
  }
}
```
→ https://i.pixeldrive.pro/t/3b9b…/04e7a6bb.png
03
Output
A URL, deterministically returned.
Same payload, same hash, same URL. Hot images come back from Redis in five milliseconds or less. Cold renders go through the headless browser pool in roughly seven hundred milliseconds. Anything past that is queue, not work.
Last 4 renders4 / 4 ok
- 04e7a6bbrendered988ms
- 136b2e32rendered1,214ms
- 2103c214cache · redis4ms
- b4a29879cache · redis6ms
04
Scale
Throughput, by adding nodes.
One container × pool size six = roughly six renders per second. Replicate the container, queue handles fan-out. The render service holds no state of its own — there is no master, no sticky session, no migration to plan.
Throughput · renders / secondcluster × pool
1 × 6
6.1
2 × 6
12.2
4 × 6
24.4
8 × 6
48.8
One environment variable per container. RENDER_CLUSTER_SIZE multiplied by RENDER_POOL_SIZE is your headroom.

Built for

Anywhere a team is opening Photoshop one too many times.

We don't pick the design. We pick up where the designer finished. Anything that goes to ten-thousand-of-it is the moment this saves you.

Paid acquisition↗
Ad creative at the rate of A/B/C/D/E…
Feed your variant matrix in once. PixelDrive renders the cross-product (placement × copy × hero image) and ships them to your ad platforms cold.
4,800 variants
per campaign brief
Marketplaces↗
Listing imagery, one per SKU
Brand-consistent product cards for every line item in a catalogue refresh, generated from the spreadsheet you already maintain.
0.7s
p50 per render
Education↗
Course covers, certificates, badges
Per-student artwork that respects the brand. Names, dates, completion status, signatures — variables, all of them.
per learner
1 cache hit thereafter
Editorial↗
Article hero images at publish time
A webhook on publish. PixelDrive composes the hero — pulled headline, byline, category palette — and writes it to your CDN before the page is live.
< 1.2s
webhook → CDN
Audio↗
Podcast episode art, by feed
Generate the cover art for every episode of every show on your network from one template. Stop opening Photoshop on Tuesdays.
no design seat
after first template
Programmatic↗
Webhooks, Zapier, n8n, your own queue
The API speaks one shape. Anything that can POST JSON can drive PixelDrive — including the next thing your team builds in-house.
1 endpoint
/v1/render

By the numbers

The metric only matters if we can show you it.

Receipts from our own load tests.
Same code that runs in production.

100concurrent
renders in 16.4s wall-clock on a single node — our own load test, repeated.
6.8rps
sustained throughput per render container before queue depth becomes the bottleneck.
≤ 5ms
cache-hit response time. 30–60% of production traffic hits cache for repeat campaigns.
24slots
per beefy container: 4 worker procs × 6 browser contexts, sized to the CPU you give it.

Pricing

Pay for renders, not seats.

Cache hits are free. Unused credits roll over. Upgrade or cancel any month.

Free

Kick the tires. Full API, 200 starter renders, no card.

$0to start

200 starter renders
Unlimited previews
API access
Community support

Start free

Basic

Steady, low-volume automation for a single product.

$12per month

2,000 renders / month
Unlimited previews
No-code integrations
Credits roll over

Start Basic

Recommended

Pro

Where most teams land. Priority API, unwatermarked output.

$29per month

8,000 renders / month
No watermark
Priority API
Custom fonts + assets

Start Pro

Business

High-volume pipelines and teams shipping at scale.

$69per month

50,000 renders / month
Dedicated throughput
Team seats
Email support

Start Business

VIP

Maximum throughput with white-glove support.

$149per month

250,000 renders / month
Top render priority
White-glove support
Custom CDN

Start VIP

Enterprise

Dedicated infrastructure, SLAs, private deployments.

Customusage-based

Unlimited renders
Dedicated render nodes
SAML + audit logs
SLA + on-call support

Talk to us

All plans use the same /v1/render endpoint. No feature gating on the API surface.

Questions

Likely already asked.

If your question is not here, mail hello@pixeldrive.pro. A real person answers within a working day.

Q.01
What does it actually output?
PNG by default. JPEG, WebP, and AVIF when the Accept header asks for them. Vector formats (SVG, PDF) are on the Studio plan.
Q.02
How fast is a single render?
Identical payloads return in 1–5 ms via the Redis cache. Cold renders average ~700 ms once we have warmed the browser pool. Bulk batches scale linearly with worker count.
Q.03
Can I bring custom fonts and assets?
Yes. Upload fonts in the editor or reference them by URL; same for image assets. Bring-your-own S3 bucket and CDN are supported on Studio.
Q.04
How is billing handled?
Credits are decremented only when a render actually runs. Cache hits are free. Unused credits roll over forever. You can upgrade, downgrade, or cancel any month.
Q.05
Is there an API, or just the dashboard?
The dashboard is a thin wrapper on the same API every customer uses. POST /v1/render for one image, /v1/jobs for batches, SSE on /v1/jobs/:id/stream for live progress.
Q.06
Can we self-host?
Scale-tier customers get a private deployment, either on our infra or theirs. We hand over the same container image we run, plus the Postgres + Redis schema and the Helm chart.

One more thing

The first thousand renders are on us.

No card. No sales call. Pull one template into the editor, mark a few fields as variables, and POST your first payload by the time the coffee is cool.

Claim 1,000 free renders See pricing

Render the images your whole campaign needs, from one template.

Bulk render

It discovers your editor.

It builds and previews — free.

It generates only when you mean it.

A template, made once.

A payload, posted in.

A URL, deterministically returned.

Throughput, by adding nodes.

Ad creative at the rate of A/B/C/D/E…

Listing imagery, one per SKU

Course covers, certificates, badges

Article hero images at publish time

Podcast episode art, by feed

Webhooks, Zapier, n8n, your own queue

Free

Basic

Pro

Business

VIP

Enterprise

The first thousand renders are on us.