Purpose
vap-showcase is positioned as a "Media Execution Control Layer for AI Agents," with the core goal of providing cost control when AI agents invoke media generation capabilities, adopting a "reserve-execute-burn/refund" billing model, and supporting pre-generation upfront quotes and budget constraints.
At the same time, this repository is the public GitHub and MCP discovery entry point for VAP Media API integration, used to let MCP-enabled clients and agent workflows connect to VAP's media tools.
Main Capabilities
- Media execution control: reserve-execute-burn/refund model, upfront quotes and budget constraints.
- Media generation: supports Flux2 Pro, Veo 3.1, Suno V5.
- Post-processing: format conversion and audio normalization based on FFmpeg.
- MCP integration: MCP endpoint https://api.vapagent.com/mcp , supports Claude Desktop, Claude Code, Cursor-compatible MCP clients, and other agent workflows that invoke VAP media tools. For clients that do not support setting request headers directly, you can use the local proxy mcp/vap_mcp_proxy.py and set VAP_API_KEY.
- Media API: a unified generation API for image, video, and music workflows. The currently public model interfaces include Pimo AI-Video, Aura Image Turbo, and Pira V5.5.
- Coding Plan API: an OpenAI-compatible API for coding agents, IDEs, editors, and automation workflows, powered by Nemesis Deep Coder, with the model ID vap-code.
- Product interface: AI Room is a conversational creation workspace where custom-trained AI agents use session context to create images, videos, voice, and music.
Usage
Developer Hub:https://vapagent.com/developer/
Media API:
- API base URL:https://api.vapagent.com/api/v1
- Create generation: POST /api/v1/generations
- Create operation: POST /api/v1/operations
- Authentication: product-scoped VAP Media API key
- MCP endpoint:https://api.vapagent.com/mcp
Coding Plan API:
- OpenAI-compatible API base URL:https://api.vapagent.com/v1
- Model ID: vap-code
- Responses endpoint: POST /v1/responses
- Chat Completions endpoint: POST /v1/chat/completions
- Authentication: product-scoped VAP Coding Plan API key
Product entry:https://vapagent.com/new API base URL:https://api.vapagent.com/api/v1
MCP integration example (Claude Desktop): the repository provides a Claude Desktop example configuration; for clients that do not support setting request headers directly, use the local proxy mcp/vap_mcp_proxy.py and set VAP_API_KEY.
Use Cases
- Agent applications that need to control costs, obtain upfront quotes, and constrain budgets in AI media generation workflows.
- Agent workflows that invoke media generation tools through MCP clients such as Claude Desktop, Claude Code, and Cursor-compatible clients.
- Workflows that need a unified generation API across images, videos, and music.
- Coding agents, IDEs, editors, and automation pipelines that need OpenAI-compatible interfaces.
- Generating images, videos, voice, and music with the help of session context in a conversational creation workspace.
Limitations
- Apart from the public model interfaces and plan names mentioned above (Lite, Pro, Max monthly Room plans), specific pricing, quotas, and quota details are not yet provided.
- The specific parameters, capability boundaries, and output specifications of each model are not yet provided.
- The specific calculation rules and implementation details of budget constraints, quotes, and the burn/refund mechanism are not yet provided.
- The specific format range supported by FFmpeg post-processing is not yet provided.
Other Information
- Website:https://vapagent.com/
- AI Room：https://vapagent.com/new
- MCP guide:https://vapagent.com/mcp.html
- Webhooks：https://vapagent.com/webhooks.html
- Status page:https://vapagent.com/status.html
- Support email: support@vapagent.com
- Model page:https://vapagent.com/models/nemesis-deep-coder.html
- Harness guide:https://vapagent.com/integrations/coding-harnesses.html
- Discovered files: server.json (MCP Registry server metadata), mcp.json (MCP package metadata), glama.json (Glama MCP listing metadata), mcp/tools.json (MCP tool schema metadata), mcp/vap_mcp_proxy.py (local stdio/http proxy for MCP clients)
- License: MIT License, see LICENSE for details
