UplinkSwitch MCP server
Official# UplinkSwitch MCP server
An [MCP](https://modelcontextprotocol.io) server for [UplinkSwitch](https://uplinkswitch.ai): every chat, embedding, rerank and transcription model on one API key, with the full price catalog readable before you sign up.
The catalog tools need **no API key**. An agent can list models, compare prices per million tokens and read a model's context window and capabilities before anyone creates an account. Only `chat` needs a key.
## Install
```bash
npx -y @uplinkswitch/mcp
```
Or from source:
```bash
git clone https://github.com/uplinkswitch/uplinkswitch-mcp
cd uplinkswitch-mcp
npm install && npm run build
node dist/cli.js
```
## Configure
Claude Desktop (`claude_desktop_config.json`), Claude Code, Cursor, Windsurf, Zed and any other MCP client:
```json
{
"mcpServers": {
"uplinkswitch": {
"command": "npx",
"args": ["-y", "@uplinkswitch/mcp"],
"env": {
"UPLINKSWITCH_API_KEY": "sk-uplink-live-..."
}
}
}
}
```
`UPLINKSWITCH_API_KEY` is only needed for `chat`. Keys come from <https://uplinkswitch.ai/console>.
`UPLINKSWITCH_BASE_URL` points the server at another host on the same network, default `https://api.uplinkswitch.com/v1`:
| Brand | `UPLINKSWITCH_BASE_URL` |
|---|---|
| [UplinkSwitch](https://uplinkswitch.ai) | `https://api.uplinkswitch.com/v1` |
| [UncensoredLLM](https://uncensoredllm.ai) | `https://api.uncensoredllm.ai/v1` |
| [RelaySwitch](https://relayswitch.io) | `https://api.relayswitch.io/v1` |
| [Unrestrictable](https://unrestrictable.ai) | `https://api.unrestrictable.ai/v1` |
| [PentestLLM](https://pentestllm.ai) | `https://api.pentestllm.ai/v1` |
The catalog tools work against every one of them without a key, so you can compare the whole network's pricing.
## Tools
| Tool | Key | What it does |
|---|---|---|
| `list_models` | no | Every model with its price per million tokens. Filter by text or by kind (`chat`, `embedding`, `rerank`, `transcription`). |
| `model` | no | One model's full entry: context window, output limit, supported parameters, endpoints, exact pricing. |
| `compare_prices` | no | Price per million tokens, cheapest input first, optionally filtered by tag such as `uncensored` or `reasoning`. |
| `chat` | yes | Run a conversation on any chat model id; returns the answer, token usage and the cost of the request. |
## Why one key for every model
UplinkSwitch is a gateway: the same key and the same base URL reach Claude, GPT, Gemini, Grok, DeepSeek, GLM, Qwen, Kimi and the open-weight models, through the OpenAI, Anthropic and Google wire formats. There is no subscription and no monthly minimum — you buy prepaid credit and pay per token at the published price.
- Prices and the full catalog: <https://uplinkswitch.ai/pricing>
- OpenAI and Anthropic compatible: `https://api.uplinkswitch.com/v1`
- The catalog this server reads: <https://api.uplinkswitch.com/v1/catalog>
- OpenAPI document: <https://api.uplinkswitch.com/openapi.json>
## Sibling servers
Each brand runs its own MCP endpoint on the same protocol. Point a client at whichever fits:
- UplinkSwitch — `https://api.uplinkswitch.com/mcp`
- UncensoredLLM — `https://api.uncensoredllm.ai/mcp`
- RelaySwitch — `https://api.relayswitch.io/mcp`
- Unrestrictable — `https://api.unrestrictable.ai/mcp`
- PentestLLM — `https://api.pentestllm.ai/mcp`
Their server cards are at `<host>/mcp/server-card` and `/.well-known/mcp.json` lists them for directory crawlers.
## License
MIT
TDQS
Scored across 4 tools
Each tool has a distinct purpose: single model lookup, chat completion, filtered catalog listing, and price comparison. However, list_models and compare_prices both surface pricing information and could be confused by an agent deciding which to call for cost data, and the bare name 'model' is less self-explanatory than the others.
Two tools follow a verb_noun pattern (list_models, compare_prices) while the other two are bare nouns (model, chat), and 'model' is a particularly vague name for a getter. The set is readable but mixes conventions without a clear rule.
Four tools is a tight, purposeful set for an LLM gateway: discovery, pricing comparison, and chat execution all earn their place. It sits at the thin end of the ideal range but nothing is redundant.
Core workflows (find a model, compare cost, run a chat) are covered with no dead ends, but obvious gateway operations are absent — no way to check account balance/credits or limits despite cost-returning chat, and no embeddings or streaming-specific entry point. Agents can work around these but will hit the gap on billing or non-chat use cases.