Generate AI images, videos, and music with deterministic pricing.
No setup required – start generating in 2 minutes:
- Create Agent → Get your API key instantly
- Add Funds → Start from $1
- Generate → Start creating images, videos, and music
VAP is an MCP server that enables image, video, and music generation directly from agentic workflows. It exposes generative media capabilities as portable, pay-per-use tools usable across Claude Desktop, Cursor, and other MCP-compatible clients.
When AI agents work with paid APIs, they need:
- Cost visibility – Know exactly what you'll pay before execution
- Retry control – Bounded, predictable retry behavior
- Clear ownership – Every task tracked and accountable
- Enterprise auth – OAuth 2.1 M2M for secure integrations
VAP provides this control layer.
VAP is an MCP Server that provides Flux image generation, Veo 3.1 video generation, and Suno V5 music generation with full cost control.
Supported AI Models:
- Flux2 Pro – Photorealistic images
- Veo 3.1 – Cinematic videos
- Suno V5 – Original music
Production Pipeline:
- FFmpeg post-processing (format conversion, audio normalization)
- Automatic quality optimization for broadcast standards
- Permanent cloud storage with instant CDN delivery
- Aspect ratio auto-detection from prompt (e.g. "16:9 landscape" → correct dimensions)
How it works:
- Pre-commit pricing – Know exact cost before execution
- Reserve → Burn → Refund – Hard budget guarantees
- Deterministic behavior – Predictable results every time
- Explicit ownership – Every task has a clear owner
Agent: "Generate an image of a sunset"
↓
VAP: "Reserving cost..."
VAP: "Reserved. Executing..."
VAP: "Success. Here's your image."
If it fails:
Agent: "Generate an image of a sunset"
↓
VAP: "Reserving cost..."
VAP: "Reserved. Executing..."
VAP: "Failed. Full refund. Error: Provider timeout"
Your agent never sees the complexity. It just gets deterministic results.
| Type | Preset |
|---|---|
| Image | image.basic |
| Video | video.basic |
| Music | music.basic |
| Campaign+SEO | streaming_campaign |
| Full Production+SEO | full_production |
All media productions are automatically normalized and delivered through a fast, orchestrated pipeline in accordance with defined broadcast quality standards.
Pricing: See vapagent.com for current pricing.
Generate an image with zero signup — no API key, no deposit:
curl -X POST https://api.vapagent.com/v3/trial/generate \
-H "Content-Type: application/json" \
-d '{"prompt": "A mountain landscape at sunset, photorealistic"}'Rate limit: 3 images per day per IP.
Ready for more? Create an agent below.
Option A: Dashboard (Recommended)
Go to vapagent.com/dashboard/signup.html and create your agent.
Option B: API
curl -X POST https://api.vapagent.com/v3/agents \
-H "Content-Type: application/json" \
-d '{"Name": "my-agent"}'Save the api_key from the response. It's shown only once.
Deposit minimum $1 to unlock generation capabilities:
curl -X POST https://api.vapagent.com/v3/deposits/init \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{"amount": 5.00, "provider": "crypto"}'Add to claude_desktop_config.json:
{
"mcpServers": {
"vap": {
"url": "https://api.vapagent.com/mcp",
"transport": "streamable-http",
"headers": {
"Authorization": "Bearer YOUR_API_KEY"
}
}
}
}Add to .cursor/mcp.json (project) or ~/.cursor/mcp.json (global):
{
"mcpServers": {
"vap": {
"command": "npx",
"args": [
"mcp-remote",
"https://api.vapagent.com/mcp",
"--header",
"Authorization: Bearer YOUR_API_KEY"
]
}
}
}Add to Cline MCP settings:
{
"mcpServers": {
"vap": {
"command": "python",
"args": ["path/to/mcp/vap_mcp_proxy.py"],
"env": {
"VAP_API_KEY": "your_api_key"
}
}
}
}Restart your client after configuration.
| Tool | What it does |
|---|---|
generate_image |
Create AI image from text prompt |
generate_video |
Create cinematic video from text (Veo 3.1) |
generate_music |
Create original music from text (Suno V5) |
upscale |
Upscale image resolution (2x/4x) |
background_remove |
Remove image background |
inpaint |
Edit specific regions of an image |
video_trim |
Trim video to specific time range |
video_merge |
Merge multiple video clips |
estimate_cost |
Preview image generation cost |
estimate_video_cost |
Preview video generation cost |
estimate_music_cost |
Preview music generation cost |
check_balance |
Check your current balance |
get_task |
Check status of a generation task |
get_operation |
Check status of an edit operation |
list_tasks |
List your recent tasks |
For environments that don't support headers, use the local proxy:
{
"mcpServers": {
"vap": {
"command": "python",
"args": ["/path/to/mcp/vap_mcp_proxy.py"],
"env": {
"VAP_API_KEY": "your_api_key"
}
}
}
}MCP Registry: io.github.elestirelbilinc-sketch/vap-e
For enterprise integrations, VAP supports OAuth 2.1 M2M (machine-to-machine) authentication via Scalekit.
OAuth Token → MCP Proxy → Validate → Resolve client_id → Agent → Execute
# 1. Get OAuth token from your identity provider
curl -X POST "https://your-tenant.scalekit.dev/oauth/token" \
-d "grant_type=client_credentials" \
-d "client_id=m2m_xxx" \
-d "client_secret=your_secret"
# 2. Link OAuth client to your VAP agent (one-time)
curl -X PUT "https://api.vapagent.com/v3/agents/me/oauth" \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{"oauth_client_id": "m2m_xxx"}'
# 3. Use OAuth token with MCP
# The MCP proxy validates tokens and maps to linked agents- SSO Integration – Connect VAP to your existing identity provider
- No API Key in Config – Tokens rotate automatically
- Audit Trail – OAuth events logged separately
pip install vap-sdkfrom vape_client import VAPEClient
client = VAPEClient(api_key="your_api_key")
result = client.generate(
description="A serene mountain landscape at sunset"
)
print(f"Image URL: {result.image_url}")
print(f"Cost: ${result.cost}")# Generate video with Veo 3.1
video = client.generate_video(
prompt="Cinematic aerial shot of coastal cliffs at golden hour",
duration=8,
aspect_ratio="16:9",
generate_audio=True
)
print(f"Task ID: {video.task_id}")
# Poll for completion
task = client.get_task(video.task_id)
print(f"Status: {task.status}")
print(f"Video URL: {task.result_url}")# Generate music with Suno V5
music = client.generate_music(
prompt="Upbeat indie folk song with acoustic guitar and warm vocals",
duration=120,
instrumental=False
)
print(f"Task ID: {music.task_id}")
# Check task status
task = client.get_task(music.task_id)
print(f"Audio URL: {task.result_url}")import asyncio
from vape_client import AsyncVAPEClient
async def main():
async with AsyncVAPEClient(api_key="your_api_key") as client:
# Image
image = await client.generate(description="A futuristic cityscape")
print(f"Image URL: {image.image_url}")
# Video
video = await client.generate_video(prompt="Ocean waves at sunset")
print(f"Video Task: {video.task_id}")
# Music
music = await client.generate_music(prompt="Lo-fi chill beats")
print(f"Music Task: {music.task_id}")
asyncio.run(main())# List recent tasks
tasks = client.list_tasks(limit=10)
for task in tasks.tasks:
print(f"{task.task_id}: {task.status} - {task.task_type}")
# Get specific task
task = client.get_task("your-task-id")
print(f"Result: {task.result_url}")| Endpoint | Method | Description |
|---|---|---|
/v3/agents |
POST | Register new agent |
/v3/agents/me/balance |
GET | Check account balance |
/v3/agents/me/oauth |
PUT | Link OAuth client (Enterprise) |
/v3/agents/me/oauth |
GET | Check OAuth link status |
/v3/tasks |
POST | Create media generation task |
/v3/tasks/{id} |
GET | Retrieve task status |
/v3/execute |
POST | Execute preset (campaign/production) |
/v3/operations |
POST | Run edit operation (upscale, bg_remove, etc.) |
/v3/operations/{id} |
GET | Check operation status |
/v3/deposits/init POST |
Initialize deposit | |
/v3/trial/generate |
POST | Free trial image (no auth) |
/mcp |
- | MCP streamable-http endpoint |
Full API Docs: vapagent.com/quick-start.html
Every task has a known cost before execution. No surprises.
Set a max budget. VAP enforces it. Hit the limit? Task rejected. Balance protected.
Every task has an explicit owner. Every failure has an address. No more "the agent did something and I don't know what."
Every output is normalized to broadcast standards. Consistent formats, predictable quality, publish-ready media. No variance between runs.
- MCP Registry: registry.modelcontextprotocol.io
- API Documentation: vapagent.com/quick-start.html
- MCP Endpoint:
https://api.vapagent.com/mcp
MIT License – see the LICENSE file for details.
VAP Media · Unified MCP Server for AI Agents
"Deterministic media production for AI agents."