> ## Documentation Index
> Fetch the complete documentation index at: https://docs.imagine.art/llms.txt
> Use this file to discover all available pages before exploring further.
# Imagine MCP
> Connect ImagineArt to Claude, Cursor, OpenClaw, Hermes, and any MCP client — no API key, billed through your existing imagine.art credits.
Imagine MCP connects ImagineArt to **Claude, Cursor, OpenClaw, Hermes**, and any client that speaks the Model Context Protocol. Use ImagineArt's full set of creative tools — image, video, music, fashion, ads, and more — with no API key, billed through your existing imagine.art credits.
By the end of setup, you can tell your agent "Generate an image of a red bicycle at sunset" and get the finished media back in the conversation.
In your client, add a custom MCP server pointing to `https://mcp.imagine.art`.
Sign in with your imagine.art account when prompted — no API key.
Ask your agent to create something. Generations draw from your existing credits.
## What is MCP?
**MCP (Model Context Protocol)** is an open standard that lets an AI agent connect to external tools. ImagineArt runs a hosted MCP server; your AI client (Claude, Cursor, and so on) acts as the host that connects to it.
When you register the ImagineArt server, its creative tools appear as native tools your agent can call. The connection is:
* **Hosted** — the server lives at `https://mcp.imagine.art`. There's nothing to run or install.
* **Authenticated by your account** — you sign in with your imagine.art login. There is no API key to generate, store, or rotate.
* **Billed through your credits** — no separate MCP pricing. The free tier includes 100 credits/day.
Two addresses, don't mix them up. `https://www.imagine.art/mcp` is the **information page** you read in a browser. `https://mcp.imagine.art` is the **server endpoint** that goes into your client's setup. Use the server endpoint for setup.
## Why Imagine MCP
Most creative MCP servers require API keys, separate billing, and credential management across platforms. Imagine MCP skips all of that.
* **No API key needed** — every request authenticates through your imagine.art account.
* **Every tool in one connection** — image, video, music, upscaling, background removal, fashion, and ad generation are all reachable through a single server, so you can chain them in one conversation.
* **Uses your current balance** — runs on the same credit system as the platform; your existing plan and balance carry over with no extra charges.
* **Zero data retention** — each request is processed independently. Prompts, outputs, and session data aren't retained.
## What you can build
Tell your agent what you're working on and it picks the right tools, chains them together, and delivers production-ready results without you leaving the conversation. Generate an image, upscale it, strip its background, then feed it into a video tool. Shoot a model in an outfit, then animate the still into a campaign clip. Turn a product photo into a finished ad — all in one agent session.
## Available tools
### Six core tools
| Tool | What it does |
| ------------------ | ------------------------------------------------------- |
| Text-to-image | Generate images from a prompt at multiple resolutions |
| Text-to-video | Generate short clips from a prompt or a reference image |
| Music generation | Produce original music or instrumentals from a prompt |
| Image upscaler | Enhance and increase the resolution of an image |
| Background remover | Cleanly strip the background from an image |
| Balance inquiry | Check your remaining credits and renewal date |
### Guided studio workflows
Beyond the six core tools, two full multi-step workflows are available — each walks your agent through a short setup, then generates a polished, production-ready result:
* **Fashion Studio** — compose a model, wardrobe, scene, and pose into a finished photoshoot, or animate a still into a campaign video. See [Fashion Studio](#fashion-studio) below.
* **Ad Studio** — turn a product and an avatar into a finished ad, image or video, using a format, hook, and setting. See [Ad Studio](#ad-studio) below.
### Specialized creative tools
Purpose-built recipes that combine the core engines into polished, ready-to-use outputs:
* **Logo generation** — clean, scalable brand marks
* **3D logo animation** — turn a flat logo into a cinematic reveal
* **Cinematic product ad** — animate a product photo into a commercial clip
* **Giant product showcase** — surreal building-scale product hero shot
* **UGC lifestyle try-on** — authentic influencer-style product photos
* **Instagram post** — scroll-stopping hero image with caption and hashtags
* **YouTube thumbnail** — high-CTR 16:9 thumbnail with overlay guidance
* **Interior design** — redesign a room from a photo or a concept
* **Jewelry video** — luxury macro product commercial
* **Cooking video** — turn a person's photo into a tutorial clip
* **Drone/aerial video** — sweeping flyover, orbit, and top-down shots
### Workflow and account helpers
* **Upload images** — bring your own reference assets into a generation
* **List generations/uploaded assets** — browse what you've made
* **Select organization and folder** — choose the workspace and where outputs are saved
## Fashion Studio
Fashion Studio is a guided, multi-step workflow for AI fashion photoshoots and fashion videos — model, wardrobe, scene, and pose composed into finished editorial or catalogue shots, or animated into short campaign clips.
Ask your agent for a fashion shoot — for example, "Generate me a Fashion post."
Your agent walks through organization → shoot type → project, then model → wardrobe → background → pose.
It generates the shoot and shows you the results. From there you can animate any still into a video.
### How it works
Fashion Studio shares one setup, then splits into two branches:
* **Shared setup:** organization → shoot type (editorial or catalogue) → project
* **Photoshoot branch:** model → wardrobe (outfit + footwear/accessories, optional) → background (optional) → pose (optional) → generate
* **Video branch:** pick a finished still → template or free-text motion → camera movement (optional) → duration/aspect/resolution → generate
| Step | What it does |
| ------------------- | ------------------------------------------------------------------------------------------ |
| Fashion model | Select an existing AI model, or create one from your own reference photos |
| Wardrobe | Select an outfit (top/bottom or dress) and, optionally, footwear or accessories |
| Background and pose | Choose a scene and pose — or shoot at a plain root with no background selected |
| Generate photoshoot | Compose model + wardrobe + scene + pose into finished photos |
| Animate to video | Turn any finished still into a short clip using an editorial or catalog template |
| Composite shoot | Merge multiple stills (for example, the same model across shots) into one reconciled frame |
Generating a photoshoot or a composite requires an active (paid) subscription. Free-tier organizations can browse and set up a project, but generation needs an upgrade.
### Example workflow
> Create an editorial lookbook shot: put the navy blazer and the white sneakers on my "Studio Model 1," in a soft daylight loft setting, 4:3.
The agent walks org → shoot type → project (reusing your existing project if you have one), then model, wardrobe, and background selection, then generates the shoot and shows the results. Ask it to "turn that into a 6-second campaign clip" afterward and it animates the still using a matching video template.
## Ad Studio
Ad Studio turns a product photo (or an existing product) into a finished ad — image or video — by chaining together a product, an avatar, a format, and an optional hook and setting.
Ask your agent to create an ad — for example, "Make a UGC-style ad for my water bottle."
It walks you through picking (or adding) a product and an avatar, then a format, hook, and setting.
Confirm resolution, aspect ratio, and duration, and it generates the ad.
### The pipeline
Ads are generated through a sequential setup: organization → marketing project → product → avatar → format → hook → setting → generate.
| Step | What it does |
| ----------------- | ---------------------------------------------------------------------------------------------- |
| Marketing project | The campaign container the ad is linked to — pick an existing one or create a new one |
| Product | Pick an existing product, or add one by uploading a photo or pasting a product URL |
| Avatar | Pick an existing AI avatar, or create one from a text prompt or reference photos |
| Format | The ad type — for example, UGC, testimonial, unboxing. Skipped if you already named a format |
| Hook and setting | Optional — an opening hook and a scene/setting for the ad |
| Generate | Compose everything into the finished ad, at your chosen resolution, aspect ratio, and duration |
Resolution, aspect ratio, and duration are all required before generating — confirm each first so credits aren't spent on an unintended output.
### Example workflow
> Create a testimonial-style ad for my ceramic mug using my "Sarah" avatar, 9:16, 1080p, 10 seconds.
The agent resolves the mug as your product and Sarah as your avatar, matches "testimonial" to its format, confirms the hook/setting, then generates the ad at the specs you gave.
## Passing parameters and references
You control each generation through a few simple parameters. You don't pass these as raw code — just describe them to your agent (for example, "make it 16:9, 4K, using the veo model") and it maps them to the right tool.
### Common parameters
| Parameter | Applies to | Notes |
| ------------ | -------------- | -------------------------------------------------------------------------- |
| Prompt | All generators | The text description of what to create |
| Model | Image/video | Each tool has a default; you can request a specific model |
| Aspect ratio | Image/video | For example, 1:1, 16:9, 9:16 — invalid values fall back to a supported one |
| Resolution | Image/video | Images: 1K/2K/4K. Video: 480p–4K (model-dependent) |
| Duration | Video | In seconds; supported lengths vary by model (commonly 4–15s) |
### Using references
* **By URL** — pass the URL of an existing image as a reference.
* **By upload** — upload your own image(s) directly and use them as references.
* **Multiple references** — for video, you can supply more than one reference image (and on some models, reference video or audio) to guide the result.
## Example workflows
Each example is just what you'd type to your agent.
**A single image**
> Generate an image of a red bicycle at sunset, 16:9, 4K.
The agent calls text-to-image and returns the finished image in the conversation.
**Chain tools into a product hero shot**
> Generate a sleek matte-black water bottle on a marble surface, then upscale it and remove the background.
Image → upscale → background removal, all in one session, ending with a transparent PNG ready to drop into a design.
**Turn your own photo into an ad**
> Here's my product photo — turn it into a 6-second cinematic ad with a luxury mood.
Upload the reference, then the cinematic product-ad tool animates it into a commercial clip.
**Build a mini brand kit**
> Create a minimal wordmark logo for "Northwind Coffee", then animate it into a 3D reveal.
Logo generation → 3D logo animation, producing both a static mark and a motion intro.
**Shoot and animate a fashion look**
> Put this jacket on a model for an editorial shoot, then turn the best still into a 6-second campaign clip.
Fashion Studio composes the shoot, then animates the chosen still using a matching video template.
**Build a UGC-style ad from a product photo**
> Here's my product — create a UGC-style ad with an avatar, 9:16, 15 seconds.
Ad Studio resolves the product and avatar, applies the UGC format, and generates the finished vertical ad.
## How to connect
### Claude
Launch Claude, then go to **Settings → Connectors → Add custom connector**.
Name it `Imagine MCP` and paste the URL: `https://mcp.imagine.art`
Click **Add → Connect**, sign in with your imagine.art account, then ask Claude to generate an image.
### Cursor
In Cursor, go to **Settings → MCP → Add new MCP server**, or edit `~/.cursor/mcp.json` directly.
Paste this configuration (the token comes from your signed-in imagine.art account):
```json theme={null}
{
"mcpServers": {
"ImagineArt": {
"type": "http",
"url": "https://mcp.imagine.art",
"headers": {
"Authorization": "Bearer "
}
}
}
}
```
Save and restart Cursor, then ask the agent to generate a hero image — it picks the right tool.
### OpenClaw
Make sure OpenClaw is installed, updated (`openclaw update`), and running. Then:
```bash theme={null}
# Add the server
openclaw mcp add imagine --url https://mcp.imagine.art --transport streamable-http --auth oauth --timeout 180 --connect-timeout 60
# Sign in (opens an authorization URL)
openclaw mcp login imagine
openclaw mcp login imagine --code 'YOUR_CODE'
# Load the new tools into running agents
openclaw mcp reload
```
**Authorizing:** `login` prints an `https://imagine.art/mcp/authorize?...` URL. Open it, approve access, then copy only the value between `code=` and `&` from the redirect (the "site can't be reached" page is normal). Wrap the code in single quotes so a stray `&` can't break the command. Codes expire in about 1–2 minutes — re-run `login` for a fresh one if needed.
```bash theme={null}
# Inspect
openclaw mcp status --verbose # list saved servers + auth state (no network call)
openclaw mcp show imagine # show this server's raw config
openclaw mcp probe imagine # connect and list tools (makes the network call)
# Maintain / reset
openclaw mcp configure imagine --timeout 180 --connect-timeout 60
openclaw mcp logout imagine # clear stored credentials (keeps the server)
openclaw mcp unset imagine # remove the server entirely
```
A healthy `status` shows `authorized` and `tokens=yes`. A probe should report the full tool list.
### Hermes and other agents (device flow)
Open a terminal and run:
```bash theme={null}
hermes mcp add ImagineArt --url "https://mcp.imagine.art"
```
For `Does this server require authentication? [Y/n]:`, type `Y` and then paste your bearer token.
For `Enable all 26 tools? [Y/n/select]:`, type `y`.
Then type `hermes` and start generating.
## Security and privacy
* **Your account is the only key** — authentication runs through your existing imagine.art login. No shared API keys, no separate credentials to secure.
* **Tokens stay local** — your client stores the OAuth token on your machine and refreshes it automatically.
* **Zero data retention** — the server processes each request independently and doesn't retain prompts, outputs, or session data.
* **Trust only your own sign-in** — never approve an authorization link or paste a code that came from anything other than your own client's login command.
## Troubleshooting
| Symptom | Cause | Fix |
| ---------------------------------------------------------- | ----------------------------------------------------------------------- | ----------------------------------------------------------------------- |
| Browser shows "This site can't be reached" after approving | Normal — nothing serves the local callback page | Ignore it; copy the `code` from the address bar |
| Login never finishes / command suspends | You included `&state=...`; the `&` backgrounded the command | Copy only up to the `&`, wrap the code in single quotes |
| `code is expired / invalid` | Auth codes expire in about 1–2 min | Re-run login, re-approve, use the fresh code |
| `Request timed out` (-32001) | The streaming connection is being dropped or slowed | Use generous timeouts; switch off VPN/proxy or try a stable network |
| Agent receives messages but never replies | Its language model is out of credits/keys | Top up or switch the agent's model |
| Ad generation fails citing an invalid format | A format name (for example, "UGC") was passed instead of its numeric id | Look up the id via the format list, or use the format picker, and retry |
| New project silently reuses an old one | Duplicate project names in the same folder | Rename projects distinctly, or select by creation date in the picker |
| Generation stalls or fails on credits | Organization credit balance exhausted | Check balance and top up, or switch organizations |
| Newly created avatar/product "not found" | Its id wasn't picked up from the picker's response | Re-select from the picker rather than reusing an old id |
**Connection vs. model:** a working ImagineArt connection only means the tools are available. Your agent still needs a working language model (with credits) to drive the conversation and decide to call those tools.
## FAQ
It uses the Model Context Protocol, an open standard that gives AI agents access to external tools. Once connected, your agent can generate images, create videos, produce music, upscale assets, remove backgrounds, run fashion photoshoots, build ads, and check your balance — all within a single conversation.
Claude, Cursor, Hermes, and OpenClaw. Any agent or client that speaks MCP can connect, including custom setups running locally or on a server.
No. Add the server URL in your agent's settings and authenticate through your imagine.art account. No keys to generate, store, or rotate.
Imagine MCP uses the same credit system as the platform. Each generation costs credits based on the tool and model selected, drawn from your existing plan. Check your balance and renewal date anytime with the balance tool.
Images typically complete in a few seconds. Videos take longer depending on duration and model. Generation runs asynchronously — your agent polls for results and delivers them the moment they're ready.
Images at multiple resolutions, short videos, and original music — all from a single prompt, plus full fashion photoshoots and ad campaigns through the studio workflows. You can chain tools in sequence within one agent session.
Yes. Fashion Studio composes a model, wardrobe, scene, and pose into finished photoshoots or short campaign videos. Ad Studio turns a product and an avatar into a finished ad, image or video, using a format, hook, and setting. Fashion Studio generation is limited to paid organizations.