promptpilot-mcp-server
# PromptPilot MCP Server
Generate images, video, and audio directly in Claude Code, Cursor, Windsurf, or any MCP-compatible AI agent.
**20+ models** — Flux, GPT-Image-1, Imagen 4, Grok Imagine, Seedance, ElevenLabs TTS, and more.
**Free models** work without an API key. Paid models require a [Pollinations](https://pollinations.ai) key.
---
## Installation
### Claude Code / Claude Desktop
```json
{
"mcpServers": {
"promptpilot": {
"command": "npx",
"args": ["-y", "promptpilot-mcp"]
}
}
}
```
With Pollinations API key (for paid models):
```json
{
"mcpServers": {
"promptpilot": {
"command": "npx",
"args": ["-y", "promptpilot-mcp"],
"env": {
"POLLINATIONS_API_KEY": "sk_your_key_here"
}
}
}
}
```
### Cursor
Add to `~/.cursor/mcp.json`:
```json
{
"mcpServers": {
"promptpilot": {
"command": "npx",
"args": ["-y", "promptpilot-mcp"]
}
}
}
```
### Windsurf
Add to `~/.codeium/windsurf/mcp_config.json`:
```json
{
"mcpServers": {
"promptpilot": {
"command": "npx",
"args": ["-y", "promptpilot-mcp"]
}
}
}
```
### VS Code (GitHub Copilot)
Add to `.vscode/mcp.json` in your workspace:
```json
{
"servers": {
"promptpilot": {
"type": "stdio",
"command": "npx",
"args": ["-y", "promptpilot-mcp"]
}
}
}
```
### Docker
```bash
docker run -i --rm \
-e POLLINATIONS_API_KEY=sk_your_key_here \
mcp/promptpilot-mcp
```
---
## Tools (8 total)
| Tool | Description |
|------|-------------|
| `list_models` | List all image/video/audio models with pricing |
| `list_styles` | List styles, lighting, camera, mood, color presets |
| `build_prompt` | Build optimized prompt from structured inputs |
| `generate_image` | Generate image (returns URL) |
| `generate_video` | Generate video (returns URL) |
| `generate_audio` | Generate speech or music (returns URL) |
| `check_balance` | Check Pollinations API balance |
| `generate_batch` | Generate up to 10 images in one call |
---
## Examples
**Generate an image:**
> "Generate a cyberpunk cityscape at night using the flux model"
**Batch generation:**
> "Generate 5 product photos with different backgrounds"
**Build + generate:**
> "Build me a portrait prompt with dramatic lighting and golden hour mood, then generate it"
**Check what's available:**
> "What video models do you have and what are their prices?"
---
## REST API
PromptPilot also offers a REST API for programmatic access from any language:
```bash
# Get your token at https://promptpilot.club/profile (REST API tab)
curl https://promptpilot.club/api/v1/generate \
-H "Authorization: Bearer pp_your_token" \
-H "Content-Type: application/json" \
-d '{"prompt": "red panda in bamboo forest", "model": "flux", "fields": "url,model"}'
```
Endpoints:
- `POST /api/v1/generate` — generate image/video/audio
- `GET /api/v1/models` — list models with prices
- `GET /api/v1/balance` — check balance
- `GET /api/v1/schema` — OpenAPI schema
- `GET /api/v1/AGENT_CONTEXT.md` — quick reference for agents
---
## Links
- [promptpilot.club](https://promptpilot.club) — Web UI
- [Pollinations AI](https://pollinations.ai) — Underlying model provider
- [npm package](https://www.npmjs.com/package/promptpilot-mcp)
- [GitHub](https://github.com/doctorm333/promptpilot-mcp-server)
## License
MIT
TDQS
Scored across 8 tools
Each tool has a clearly distinct purpose with no overlap: build_prompt constructs prompts, check_balance checks API credits, generate_* tools create different media types, and list_* tools provide metadata. The separation between generate_image, generate_batch, generate_video, and generate_audio is unambiguous, and list_models vs list_styles serve different informational needs.
All tools follow a consistent verb_noun naming pattern with snake_case throughout. The verbs are descriptive and appropriate: build_, check_, generate_, and list_ are used consistently across related tools, making the set predictable and easy to understand at a glance.
With 8 tools, this server is well-scoped for its purpose of interacting with the Pollinations API for media generation and prompt building. Each tool earns its place by covering distinct aspects: prompt construction, balance checking, media generation (with batch optimization), and model/style listing, without being overwhelming or insufficient.
The tool set covers core workflows comprehensively: prompt building, credit management, media generation (image, video, audio, batch), and resource listing. A minor gap is the lack of tools for managing generated content (e.g., delete or retrieve metadata), but agents can work around this using the provided URLs and API balance checks.