Media-infrastructure
<p align="center">
<img src="assets/logo.png" alt="VAP AI Saraswati emblem" width="120">
</p>
# VAP AI
Agent-native AI platform for AI Room, Media API, and Coding Plan API.
[Website](https://vapagent.com/) | [Developer Hub](https://vapagent.com/developer/) | [AI Room](https://vapagent.com/new) | [Dashboard](https://vapagent.com/new-dashboard/) | [Status](https://vapagent.com/status.html)
## Products
### AI Room
Conversational creative workspace where custom-trained AI agents use session context to create images, video, voice, and music.
- Product URL: https://vapagent.com/new
- Plans: Lite, Pro, and Max monthly Room plans
### Media API
Unified generation API for image, video, and music workflows. Current public model surfaces include Pimo AI-Video, Aura Image Turbo, and Pira V5.5.
- Developer Hub: https://vapagent.com/developer/
- 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 for coding agents, IDEs, editors, and automation workflows. Powered by Nemesis Deep Coder with model ID `vap-code`.
- Developer Hub: https://vapagent.com/developer/
- Model page: https://vapagent.com/models/nemesis-deep-coder.html
- Harness guide: https://vapagent.com/integrations/coding-harnesses.html
- 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
## Start From The Product Surface
Use these entry points for new integrations:
| Need | Link |
| --- | --- |
| Use AI Room | https://vapagent.com/new |
| Generate a Media API key | https://vapagent.com/developer/?key=media#keys |
| Generate a Coding Plan API key | https://vapagent.com/developer/?key=code#keys |
| View Coding Plan API plans | https://vapagent.com/new-dashboard/?billing=code#plans |
| Read Developer Hub | https://vapagent.com/developer/ |
| See Nemesis Deep Coder | https://vapagent.com/models/nemesis-deep-coder.html |
## Media API Via MCP
This repository remains the public GitHub and MCP discovery surface for VAP Media API integrations. MCP is still supported for Claude Desktop, Claude Code, Cursor-compatible MCP clients, and other agent workflows that call VAP media tools.
Claude Desktop example:
```json
{
"mcpServers": {
"vap": {
"url": "https://api.vapagent.com/mcp",
"headers": {
"Authorization": "Bearer YOUR_MEDIA_API_KEY"
}
}
}
}
```
For clients that do not support headers directly, use the local proxy in `mcp/vap_mcp_proxy.py` and set `VAP_API_KEY`.
## API Examples
### Media API
```bash
curl -X POST https://api.vapagent.com/api/v1/generations \
-H "Authorization: Bearer YOUR_MEDIA_API_KEY" \
-H "Content-Type: application/json" \
-d '{"kind":"image","prompt":"a neon city at night"}'
```
### Coding Plan API
```bash
curl -X POST https://api.vapagent.com/v1/chat/completions \
-H "Authorization: Bearer YOUR_CODING_PLAN_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"vap-code","messages":[{"role":"user","content":"Review this patch."}]}'
```
## Current Product Contracts
Use these current public contracts from Developer Hub:
- Media API: `https://api.vapagent.com/api/v1`
- Media generation: `POST /api/v1/generations`
- Media operations: `POST /api/v1/operations`
- MCP: `https://api.vapagent.com/mcp`
- Coding Plan API: `https://api.vapagent.com/v1`
## Discovery 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
## Links
- Website: https://vapagent.com/
- Developer Hub: https://vapagent.com/developer/
- AI Room: https://vapagent.com/new
- MCP guide: https://vapagent.com/mcp.html
- Webhooks: https://vapagent.com/webhooks.html
- Status: https://vapagent.com/status.html
- Support: support@vapagent.com
## License
MIT License. See [LICENSE](LICENSE).
TDQS
Scored across 16 tools
Most tools have distinct purposes, but there is notable overlap between get_operation and get_task, which both retrieve task/operation statuses, potentially causing confusion. Additionally, the three estimate_* tools are clearly differentiated by media type, but their naming and purpose similarity could lead to misselection if not carefully read.
The naming is mostly consistent with a verb_noun pattern (e.g., generate_image, check_balance, list_tasks), but there are minor deviations like ai_edit (noun_verb) and background_remove (noun_verb). These inconsistencies are few and do not severely impact readability, but they break the overall pattern.
With 16 tools, the count is reasonable for a media infrastructure server covering image, video, and music generation and editing, plus account management. It is slightly on the higher side but well within a manageable scope, as each tool serves a specific function in the domain.
The tool set provides good coverage for media generation and editing, including CRUD-like operations (generate, edit, upscale, list, get status) and cost estimation. However, there are minor gaps, such as no explicit tool for deleting tasks or managing media assets beyond generation and editing, which agents might need to work around.