unorouter-mcp
by unorouter
README.md
# unorouter-mcp
MCP server for [UnoRouter](https://unorouter.com), an open-source OpenAI-compatible gateway serving 200+ models (Claude, GPT, Gemini, DeepSeek, Kimi, GLM, Qwen and more) behind a single API key. Models with a `:free` suffix cost nothing.
Gives any MCP client (Claude Desktop, Claude Code, Cursor, Windsurf, and friends) three tools:
- `search_models`: filter the live catalog, optionally free models only
- `get_pricing`: USD per-1M-token list prices from the public pricing endpoint
- `chat`: send a prompt to any model and get the reply with token usage
## Setup
Create an API key at [unorouter.com/token](https://unorouter.com/token) (Discord or GitHub sign-in, no card required).
### Claude Desktop / any MCP client
```json
{
"mcpServers": {
"unorouter": {
"command": "npx",
"args": ["-y", "unorouter-mcp"],
"env": {
"UNOROUTER_API_KEY": "sk-your-key"
}
}
}
}
```
### Claude Code
```bash
claude mcp add unorouter -e UNOROUTER_API_KEY=sk-your-key -- npx -y unorouter-mcp
```
## Environment
| Variable | Required | Default |
| --- | --- | --- |
| `UNOROUTER_API_KEY` | yes | none |
| `UNOROUTER_BASE_URL` | no | `https://api.unorouter.com` |
## Notes
- Free models run on shared pools with a light per-model rate limit (about 1 request per minute per model); rotate between free models or use paid ones for throughput.
- The whole UnoRouter stack is open source: [github.com/unorouter](https://github.com/unorouter).
## Development
```bash
bun install
bun run build
UNOROUTER_API_KEY=sk-... node dist/index.js
```
TDQS
A4.2/5.0
Scored across 3 tools
Disambiguation5/5
Each tool targets a distinct concern: model discovery, pricing lookup, and chat completion. There is no overlap or ambiguity between them.
Naming Consistency4/5
Two tools follow the verb_noun pattern (search_models, get_pricing), while 'chat' is a bare verb. This is a minor deviation but the set remains readable and predictable.
Tool Count5/5
Three tools is appropriately scoped for a model-routing/chat API. Each tool has a clear purpose and none feel redundant or missing.
Completeness4/5
The surface covers the essential workflows: discover models, check pricing, and send chats. Minor gaps like streaming or detailed model metadata exist but are not critical for the core use case.
Maintenance
ActivityMaintained
ResponsivenessNo issues