What exactly is an MCP server?

Model Context Protocol is the open standard AI clients use to discover and call external tools. Point a client at this server once and your agent gains eight image tools it can call on its own — no plugin code, no SDK per provider.

Which agents and IDEs can connect?

Anything that speaks MCP: Claude Desktop and Claude.ai, Cursor, VS Code, Copilot, Windsurf, Zed, Codex, Antigravity, Goose, OpenCode, Cline, Roo and more. There is also an agent skill (npx skills add imagemcp/skill) and a plain HTTP API if your stack doesn't speak MCP.

Do I need my own OpenRouter or Fal.ai keys?

No. One imagemcpserver key covers every model behind the server, and you get a single bill instead of an account per provider. Bring-your-own-key is supported on the Enterprise plan if you'd rather pay providers directly.

Each tool call costs a fixed number of credits — 1 for a format conversion, 5 for a vector SVG, 8 to 30 for a generation depending on the model tier. Your plan refills credits monthly, and credit packs top you up without expiring. The Free plan starts you with 30 credits a month, no card required.

Can I call it as a normal REST API?

Yes. Every MCP tool has an HTTP equivalent that takes the same arguments and a Bearer token, so you can use it from a backend job or CI pipeline that has no agent attached.

Where do generated images live?

Outputs are uploaded to object storage and returned as a public URL by default, which keeps agent context windows small. Pass responseFormat: "b64_json" on any tool if you'd rather receive the raw bytes and never have the file hosted.

What does a request log actually record?

Prompt, model, latency, HTTP status and the exact credits deducted, per call and per key. It's the audit trail you need when an autonomous agent is spending on your behalf.

Can I cancel or change plan later?

Any time, from the billing tab in your dashboard. Cancelling keeps your access until the end of the period you already paid for, and credit packs you bought stay in your balance.

For AI that builds websites

Help your AI create
websites people remember.

Not generic. Not forgettable. Not AI slop.

Your agent can already write the markup. It still leaves grey boxes where the hero shot, the product cutout and the OG card should be. This is the visual layer that fills them — from the editor you already build in.

Start building See examples

No card needed · 30 credits every month

Memorable
visuals
Cleaner
brand feel
Better first
impressions
Built for
developers

Made with ImageMCP

Twelve ways an idea can look finished.

Scroll to explore →

Forma HouseArchitecture studio

Arc JournalIndependent magazine

Ember & OakCoffee roasters

Juniper TableNeighbourhood restaurant

Moss RitualBotanical skincare

Aster RunSneaker studio

Casa BrisaBoutique hotel

Petal & StemFlower studio

Kindred CareWellness clinic

Forge AthleticsMovement studio

Morrow ClayCeramics shop

After HoursCreative agency

Forma HouseArchitecture studio

Arc JournalIndependent magazine

Ember & OakCoffee roasters

Juniper TableNeighbourhood restaurant

Moss RitualBotanical skincare

Aster RunSneaker studio

Casa BrisaBoutique hotel

Petal & StemFlower studio

Kindred CareWellness clinic

Forge AthleticsMovement studio

Morrow ClayCeramics shop

After HoursCreative agency

Generated visual directions for the sites your agent is already building.

One server, eight tools

Everything a page needs, from brief to build folder

Draw it, cut it out, sharpen it, shrink it, ship it. The whole pipeline arrives with one config line — no per-provider SDKs, no separate keys, no design handoff.

Generate Image

Text to image with 40+ AI models

8–30 creditsgenerate_image

Edit / Inpaint

Edit, inpaint or restyle any image

8–30 creditsedit_image

Remove Background

Cut the subject out in one call

8 creditsremove_background

Upscale Image

Crisp super-resolution in one call

15 creditsupscale_image

Transparent PNG

Generate and cut out in one call

16–38 creditsgenerate_transparent_image

Text to SVG

Convert text to clean, editable SVG

5 creditstext_to_svg

Compress Image

Re-encode smaller with a quality dial

1 creditcompress_image

Convert Format

PNG, JPG, WebP, GIF, TIFF, AVIF

1 creditconvert_format

How it works

Five minutes to set up.
Then it just answers.

1
Get your API key
Create your account and generate a key. Free tier, no card.
2
Point your editor at it
Paste the MCP block into Cursor, Claude Code or Copilot — or call the REST API from your build step.
3
Ask for the artwork
Describe the image in the same breath as the component. Production-ready files come back.
4
Keep an eye on spend
Every call logged: prompt, model, latency and credits, per key.

MCP client config

{  "mcpServers": {    "imagemcp": {      "command": "npx",      "args": [        "-y",        "mcp-remote",        "https://mcp.imagemcpserver.com/mcp",        "--header",        "x-api-key:sk-img-gen-your_key_here"      ]    }  }}

Works in Claude Code, Cursor, Windsurf, Copilot, Codex, Cline — anywhere an mcpServers block is accepted.

AI agent or app

ImageMCPServer

Generate Image
Edit / Inpaint
Remove Background
Upscale Image
Transparent PNG
Text to SVG
Compress Image
Convert Format

REST API

No agent? No problem.
It is just HTTP.

Every tool an MCP client can call is also a plain JSON endpoint. Drop it into a cron job, a Rails controller, a Lambda — anywhere you can make a POST request.

One key, both surfaces: The same API key and the same credit balance power your MCP client and your backend. Nothing to reconcile.
Versioned and typed: Everything lives under /v1. Failures carry an error.type and error.code you can branch on, so retries need no string matching.
OpenAPI, served live: Generate a client, import into Postman, or hand the spec to an agent. It comes from the running server, so it cannot drift.
Honest limits: 60 requests a minute per key, with X-RateLimit headers on every response and Retry-After on a 429.

Read the API reference OpenAPI spec

curl -X POST https://api.imagemcpserver.com/v1/generate \  -H "Authorization: Bearer $IMAGEMCP_API_KEY" \  -H "Content-Type: application/json" \  -d '{    "prompt": "A ceramic coffee cup on a walnut desk, morning light",    "aspectRatio": "16:9"  }'

200 OK

{  "success": true,  "result": {    "imageUrl": "https://cdn.imagemcpserver.com/gen/8f2a.png",    "model": "google/gemini-2.5-flash-image",    "latency": "4.21s"  },  "deductedCredits": 15,  "userCredits": 2485}

See it work

Watch an agent build a page with real artwork

72 seconds, unedited: a coding agent replaces every grey placeholder with images it generates, cuts out and upscales through this server.

generate_imageremove_backgroundedit_imageupscale_image+ 4 more — every call in your request log

Run the same tools yourself

Every image a website needs

The six jobs that stall every launch

Layout is the easy part. What holds a site back is the artwork — and it is the part your agent currently hands back as a placeholder.

Landing page heroes

The one image above the fold that decides whether anyone scrolls. Generated to your brief, in your aspect ratio.

See how

Open Graph cards

Every route gets a share image instead of a blank rectangle in Slack, iMessage and X.

See how

Blog & article covers

A cover per post, on publish, from the title you already wrote — no stock library subscription.

See how

Product & store images

Cut the background off a supplier photo, drop it on your own surface, and ship a catalogue that matches.

See how

Icons & favicons

One mark, vectorised and exported at every size a browser, PWA manifest and app store asks for.

See how

Social graphics

Launch posts, feature announcements and ad creative in the same visual language as the site.

See how

Actual Output

Not a mockup — actual output

Four unedited generate_image calls, with the model, latency and seed each one actually returned.

Illustrated golden valley at sunset with layered mountain ridges and a winding river — gemini-2.5-flash-image · 7.04s · seed 679263
“A golden valley at sunset, layered mountain ridges receding into warm haze, winding river”

Neon-lit cyberpunk cityscape at night with mist drifting between skyscrapers — gemini-2.5-flash-image · 7.89s · seed 565146
“A busy cyberpunk cityscape at night, neon-lit skyscrapers, drifting mist, cinematic”

Product shot of a smartwatch on cream fabric projecting a glowing holographic clock face — gemini-2.5-flash-image · 7.01s · seed 106312
“Close-up product photograph of a smartwatch displaying a glowing holographic time interface”

Abstract composition of overlapping golden circles and arcs outlined in black on cream paper — gemini-2.5-flash-image · 7.93s · seed 53382
“Overlapping golden circles, arcs and quarter-rounds with crisp black outlines on cream paper”

Generate your own

Simple pricing

Cheaper than one stock photo subscription

Prepaid credits, one bill for every model, and no per-designer seat to buy. No minimum commitment — cancel from the dashboard whenever you like.

Free

$0/month

30 credits / month

30 Credits / month
Basic image editing tools
Text → Image
API & MCP Server access

Get started free

Starter

$20/month

1,000 credits / month

1,000 Credits / month
Bring your own OpenRouter key
All image editing tools
Text → Image
Text → SVG
Image → SVG
API & MCP Server access

Start Starter

Pro

$50/month

3,000 credits / month

3,000 Credits / month
Everything in Starter
Bring your own OpenRouter key
Priority processing
Early access to new tools

Start Pro

Enterprise

$100/month

8,000 credits / month

8,000 Credits / month
Everything in Pro
Bring your own OpenRouter key
Maximum rate limits & Priority support

Start Enterprise

FAQ

Questions before you wire it in

Still stuck? Mail us at hello@imagemcpserver.com and a human will answer.

MCP endpoint
https://mcp.imagemcpserver.com/mcp

It gives whatever tool you already build in — Cursor, Claude Code, Copilot, or your own deploy script — the ability to produce the artwork a page needs: hero images, Open Graph cards, product cutouts, favicons, blog covers. Instead of pausing the build to open a stock site or a design tool, you ask for the image in the same sentence as the component and get a production-ready file back.

Build the version people remember

Wire it into Claude Code, Cursor, Copilot or your own build step in about five minutes, and stop shipping pages with the placeholders still in them.

Start building free Read the docs

No credit card required. Free tier includes 30 credits every month.

Help your AI createwebsites people remember.

Twelve ways an idea can look finished.

Everything a page needs, from brief to build folder

Generate Image

Edit / Inpaint

Remove Background

Upscale Image

Transparent PNG

Text to SVG

Compress Image

Convert Format

Five minutes to set up.Then it just answers.

Get your API key

Point your editor at it

Ask for the artwork

Keep an eye on spend

No agent? No problem.It is just HTTP.

Watch an agent build a page with real artwork

The six jobs that stall every launch

Landing page heroes

Open Graph cards

Blog & article covers

Product & store images

Icons & favicons

Social graphics

Not a mockup — actual output

Cheaper than one stock photo subscription

Free

Starter

Pro

Enterprise

Questions before you wire it in

I build websites, not agents — what does this actually do for me?+

What exactly is an MCP server?+

Which agents and IDEs can connect?+

Do I need my own OpenRouter or Fal.ai keys?+

How do credits work?+

Can I call it as a normal REST API?+

Where do generated images live?+

What does a request log actually record?+

Can I cancel or change plan later?+

Build the version people remember

Help your AI create
websites people remember.

Five minutes to set up.
Then it just answers.

No agent? No problem.
It is just HTTP.