# avotsai/avots-mcp [Health: Active]

**Category:** 🔗 Aggregators  
**Repository:** https://github.com/avotsai/avots-mcp  
**GitHub Stars:** 0  
**Views:** 3  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/avotsai-avots-mcp

## Description
Hosted multi-model AI media + chat: image, video, audio, face-swap and talking-avatar generation plus chat across 300+ models (Claude, GPT, Gemini, DeepSeek…) from one balance and one key.

## Claude Desktop Quick Installation
Heuristic fallback — verify the package name and runner against the repository README before running it. Uses `npx` (confidence: low):

```json
"mcpServers": {
  "avots-mcp": {
    "command": "npx",
    "args": ["-y","avotsai-avots-mcp"]
  }
}
```

## Documentation

## What avotsai/avots-mcp MCP server does

The avotsai/avots-mcp MCP server exposes avots.ai’s multi-provider AI platform through nineteen MCP tools. It combines chat models from providers such as Claude, GPT, Gemini, DeepSeek, and Sonar with media generation services for images, video, audio, music, narration, talking avatars, and face-based creative effects.

The server also includes utility and workflow tools. Agents can inspect the current balance, subscription tier, and pricing catalog; list available models, saved avatars, and studio trends; poll asynchronous jobs; and create calendar events from natural-language instructions. Calendar output can be sent to a linked Apple or Google calendar or returned as an ICS download.

## How it works

The hosted endpoint is `https://mcp.avots.ai/` and communicates using the MCP `2025-06-18` specification over Streamable HTTP. Most clients send `Authorization: Bearer av_mcp_…` with the MCP key created in avots.ai. Claude.ai web uses an OAuth connection flow instead of requiring the user to paste a token.

Calls draw from the same avots balance used by the web application and Telegram bot. Balance and model-listing operations are free, while generation and chat operations have per-call costs. Prices can vary by model, so an agent can call `list_models` before selecting a model.

Video-related operations use a confirmation step to reduce accidental spending. The initial request returns a preview with cost and alternatives; the client submits the job only after the user confirms a model and sends the request again with confirmation enabled. Video, audio, and some other generation tasks may run asynchronously and can be checked with `check_job`.

## Setup and configuration

Create an account at avots.ai, then generate an MCP key from Settings → Integrations. Configure the MCP client with the hosted endpoint and the authentication method documented for that client. The repository includes setup guides and ready-to-paste configuration examples for Claude Desktop, Cursor, and Cline, as well as instructions for other MCP clients.

Claude Desktop, Cursor, Cline, and other bearer-token clients may use `mcp-remote` in their client configuration. That utility is a client-side connection method, not the avots server itself. Claude.ai web connects through its OAuth flow, while Claude Code uses its MCP configuration command.

## Tools and capabilities

The documented tool set supports:

- Sending prompts to supported chat models.
- Generating or editing images, including optional native SVG output through Recraft.
- Generating videos, editing scenes, swapping faces, and re-voicing clips with lip-sync.
- Creating reusable face identities, talking-avatar videos, and vertical AI-vlog clips.
- Producing music and TTS narration, including cloned-voice narration where supported.
- Creating photo montages, vintage travel posters, and trend-based face templates.
- Listing models, prices, avatars, and trends before starting paid work.
- Polling asynchronous jobs and creating calendar events.

The avotsai/avots-mcp MCP server is suited to workflows that chain multiple tools, such as producing an image, animating it, and adding a music track through one account.

## Limitations and notes

Tool calls are billed against the avots balance, and daily USD caps and per-key rate limits apply. Generation costs vary widely: the README gives approximate ranges from about 200 tokens for some images to about 5,000 tokens for a 10-second 1080p Kling Pro video. The live model catalog is the authoritative source for current prices.

The avotsai/avots-mcp MCP server does not remove provider-specific constraints or guarantee synchronous results. Video generation is asynchronous and requires explicit confirmation, while model availability and pricing can change. A valid avots account, MCP key, and sufficient balance are needed for billed operations.

_Full upstream README: https://allmcps.com/mcp/avotsai-avots-mcp/readme_