> ## Documentation Index
> Fetch the complete documentation index at: https://docs.budgetpixel.com/llms.txt
> Use this file to discover all available pages before exploring further.
# MCP Server
> Connect Claude, ChatGPT, Cursor, or any MCP-compatible agent to BudgetPixel and generate images, video, and music from the conversation — no API key needed.
BudgetPixel runs a remote [Model Context Protocol](https://modelcontextprotocol.io) server
that gives AI agents tools to create with your BudgetPixel account: generate **images,
video, and music**, upload reference files, check pricing and credits, and publish
finished work to the BudgetPixel community — all from the chat.
```
https://mcp.budgetpixel.com/mcp
```
There are no API keys involved: you sign in with Google during the connect flow, and the
connector acts as your account. Plan credits, model discounts, and free-generation
allowances apply exactly as they do on budgetpixel.com.
Generation, uploads, and posting through MCP require the **Premium**, **Pro**, or
**Ultra** plan — the same plans that include the developer API. Any account can connect
and browse the model catalog; agents on other plans get a clear upgrade notice when they
try to generate.
BudgetPixel exposes **two** MCP endpoints. This page documents the **product server** at
`mcp.budgetpixel.com/mcp` — the one that creates images, video, and music with your
account. Separately, this documentation site hosts a **docs-search server** at
`docs.budgetpixel.com/mcp`, giving agents read-only search tools over these docs —
useful alongside the API when you want your agent to answer integration questions from
the documentation itself.
## Connect your client
1. In [claude.ai](https://claude.ai) (or the desktop app), open **Settings → Connectors**
and click **Add custom connector**.
2. Name it `BudgetPixel` and paste the server URL:
```
https://mcp.budgetpixel.com/mcp
```
3. Click **Add → Connect** and sign in with Google. Then ask Claude to
*"generate an image (or video, or song) of … using BudgetPixel"*.
1. Register the connector with one command:
```bash theme={null}
claude mcp add --transport http budgetpixel https://mcp.budgetpixel.com/mcp
```
2. Run `/mcp` in Claude Code, select **budgetpixel**, and choose **Authenticate** —
sign in with Google.
3. Ask Claude Code for an image, video, or music track — it calls the `budgetpixel`
tools and saves results to your account.
1. Open **Settings → Apps → Advanced settings** and turn on **Developer mode**.
2. Under **Apps**, click **Create app**, name it `BudgetPixel`, and paste the server
URL as the MCP connection:
```
https://mcp.budgetpixel.com/mcp
```
3. Click **Create**, then connect and sign in with Google. Ask ChatGPT to generate
with BudgetPixel.
ChatGPT connects MCP servers as "apps" via Developer mode (Settings → Apps) — this is
separate from ChatGPT's built-in Connectors, and requires a ChatGPT plan with
Developer mode enabled.
1. Open **Cursor Settings → Tools & Integrations** and click **New MCP Server** —
this opens your `mcp.json`.
2. Add BudgetPixel as a remote server by URL:
```json theme={null}
{
"mcpServers": {
"budgetpixel": { "url": "https://mcp.budgetpixel.com/mcp" }
}
}
```
3. Save — Cursor opens your browser to sign in with Google. In the agent panel, ask
for an image, video, or track.
Any client that speaks remote MCP over **streamable HTTP** connects with the one URL:
```
https://mcp.budgetpixel.com/mcp
```
The server implements standard OAuth 2.1 with PKCE and dynamic client registration,
so your agent handles the Google sign-in flow automatically — no manual token
configuration.
## Available tools
| Tool | What it does |
| ------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `generate_image` | Text-to-image, image-to-image, and multi-reference editing across the full image catalog. Results appear directly in the chat. |
| `generate_video` | SeeDance 2.0 and SeeDance 2.0 Mini — text-to-video, image-to-video, reference-images-to-video, and video-to-video editing. Priced per second by resolution; the tool echoes the exact cost before results arrive. |
| `generate_music` | Music 3.0 (default), Music 2.6, Lyria 3, and Mureka V9 — full songs with vocals from your lyrics, or instrumentals from a style prompt. Flat price per track. |
| `check_generation_status` | Poll a pending video or music job (agents do this automatically) and fetch finished results. |
| `get_generation_history` | Your recent generations — images, videos, or music — with links that can be reused as inputs for edits or posts. |
| `get_credit_balance` | Current credit balance for the signed-in account. |
| `list_models` | The live model catalog with prices and capabilities. Works without signing in. |
| `upload_media` | Upload a local image, video, or audio file (up to 50 MB) to use as a generation input. |
| `create_social_post` | Publish finished images, a video, or a music track (with cover art, genre, and lyrics) to your public BudgetPixel feed — only when you explicitly ask. |
## Billing
Generations through MCP spend the **same credits** as the BudgetPixel studio:
* Your plan's monthly credits, model discounts, and the Pro/Ultra daily free-models
allowance apply automatically.
* Images are priced per image, video per second by resolution, and music per track —
the same published prices as the [API](/concepts/pricing-and-credits). Your agent can
check any price up front with `list_models`.
* Failed generations are never charged.
* Publishing a community post costs **10 credits** (spam prevention — same policy as the
API), and posts are limited to 20 per hour. Posts created this way carry a
**"via MCP"** label for transparency.
## Try asking
Once connected, prompts like these exercise the full toolchain:
* *"Generate a cinematic portrait of a red fox in falling snow — give me 3 options"*
* *"Animate the best one into a 5-second 720p video, gentle camera push-in"*
* *"Write and sing a warm folk song about two foxes in the snow"*
* *"Post the song to my BudgetPixel feed with the portrait as cover art"*
Everything your agent creates is saved privately to your BudgetPixel account — images
and videos land in your workshop history, music in your audio studio — so nothing is
lost when the chat ends.
## MCP or the REST API?
For AI agents and chat — Claude, ChatGPT, Cursor. OAuth sign-in, no keys, a human in
the loop. Best when *you* are creating, with an agent doing the driving.
For applications — API keys, per-model endpoints, full parameter control, webhooks-free
polling. Best when your *software* is creating on a schedule or at volume.
Both surfaces expose the same models at the same credit prices, and both label their
creations (`via API` / `via MCP`) in your generation history.
## FAQ
No. You sign in with your BudgetPixel account (Google SSO) during the connect flow —
no keys to create or paste. API keys belong to the [developer API](/authentication),
which is aimed at building applications.
Premium, Pro, and Ultra. Any account can connect and use `list_models`; generation,
uploads, and posting require an eligible plan. Manage your plan at
[budgetpixel.com/subscription](https://budgetpixel.com/subscription).
The same catalog as the API — Seedream, FLUX, GPT Image 2, Nano Banana, Qwen and more
for images; SeeDance 2.0 / 2.0 Mini for video; Music 2.6, Lyria 3, and Mureka V9 for
music. The catalog updates automatically as models launch; `list_models` always
returns the live list with prices.
Everything is saved privately to your account, exactly like studio generations —
check your workshop history or audio studio, or ask the agent for
`get_generation_history`. Media links returned in chat are time-limited; your account
copies are permanent.
No — the posting tool is instructed to publish only when you explicitly ask, posts
cost 10 credits each, and every post is labeled "via MCP" on the feed.