# RouterHub Quickstart

Uncensored and abliterated AI models — GLM, DeepSeek, Qwen and Kimi variants plus GPT, Claude, Gemini and Grok — through one API key. OpenAI-compatible and Anthropic-compatible APIs. Use each model’s supported API. Pay per token. Requests use the API protocol listed for the selected model. Create an account at https://routerhub.si/register, add credits from https://routerhub.si/dashboard/billing, and create a key at https://routerhub.si/dashboard/keys. API keys are shown once. Keep them private; do not ask a user to paste one into a public chat, source repository, or browser code.

## Discover models

Fetch https://routerhub.si/models.json for the current model IDs, supported protocols, and published rates. Select a model ID exactly as listed. Do not infer protocol support or model limits from the model name. Responses and Chat Completions use a bearer API key; Anthropic Messages uses x-api-key and anthropic-version: 2023-06-01.

Rates are USD per million tokens, rather than a fixed price per request. Estimated request cost is input tokens / 1,000,000 × inputPrice + output tokens / 1,000,000 × outputPrice; reported cache usage and any provider-specific charges can change the final billed amount. An omitted cachedInputPrice is not a published cached input rate. View usage charges at https://routerhub.si/dashboard/usage.

## Tool compatibility

OpenCode supports the published Responses, Anthropic Messages, and Chat Completions models. Continue supports compatible models on these APIs; use OpenCode or Codex for codex-auto-review. Cline supports Chat Completions and Claude models available in its Anthropic model picker; choose OpenCode or Continue for a Claude model missing from that picker. Codex requires Responses; Claude Code requires Anthropic Messages. DeepSeek, GLM, and Grok currently use Chat Completions, so choose OpenCode, Continue, Cline, or a Chat Completions SDK for those models. Each setup guide and MCP list_models with a tool filter returns only compatible model IDs. Request model-specific setup context from https://routerhub.si/agent-setup.md?tool=opencode&model=EXACT_MODEL_ID, or call the read-only MCP get_setup_guide tool with the matching tool and model.

## API base

```text
https://routerhub.si/v1
```

## Chat Completions API: DeepSeek, GLM, and Grok

Use https://routerhub.si/v1/chat/completions with a model whose models.json protocols contains chat. Current published Chat Completions IDs: `grok-4.7`, `deepseek-v4-pro`, `deepseek-v4-pro-uncensored`, `deepseek-v4.1-flash`, `deepseek-v4.1-flash-uncensored`, `glm-5.3-flash`, `glm-5.3-flash-uncensored`, `glm-5.3`, `glm-5.3-uncensored`, `gemini-4-argon`, `glm-5.3-flash-abliterated`, `glm-5.3-abliterated`, `qwen3.8-max`, `qwen3.8-max-uncensored`, `kimi-k3.1-pro`, `kimi-k3.1-pro-uncensored`, `deepseek-v4.1-flash-abliterated`, `deepseek-v4-pro-abliterated`, `deepseek-v4.1-flash-1017-abliterated`. Use the Chat Completions client method in your app.

```bash
curl --connect-timeout 30 --max-time 600 https://routerhub.si/v1/chat/completions \
  -H "Authorization: Bearer $ROUTERHUB_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"deepseek-v4.1-flash","messages":[{"role":"user","content":"Reply with OK."}]}'
```

Replace the example model ID with another exact Chat Completions ID from the current catalogue. Grok uses the RouterHub base URL and RouterHub API key; an xAI API key is not needed. Check current availability and rates before sending a request. SDK examples and current input, output, and cached input rates:

- DeepSeek: https://routerhub.si/guides/deepseek
- GLM: https://routerhub.si/guides/glm
- Grok: https://routerhub.si/guides/grok

## Responses API: GPT and Codex

https://routerhub.si/v1/responses. Use only a model whose models.json protocols contains responses.

```bash
curl --connect-timeout 30 --max-time 600 https://routerhub.si/v1/responses \
  -H "Authorization: Bearer $ROUTERHUB_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-6.1-sol","input":"Reply with OK."}'
```

## Anthropic Messages API: Claude

https://routerhub.si/v1/messages. Use only a model whose protocols contains messages.

```bash
curl --connect-timeout 30 --max-time 600 https://routerhub.si/v1/messages \
  -H "x-api-key: $ROUTERHUB_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{"model":"claude-opus-5-5","max_tokens":64,"messages":[{"role":"user","content":"Reply with OK."}]}'
```

## Agent instructions

An AI agent may identify a suitable model, explain setup, and generate a user-specific configuration. The human account owner must sign up, add credits, create the API key, and enter the secret directly into their local tool. Never request, receive, store, or commit the user's password or API key. Before making a paid inference request, get the user's confirmation and clearly state it will consume prepaid credits. Verify a successful request in the Usage page.

## Tool guides

- [OpenCode](https://routerhub.si/guides/opencode)
- [VS Code with Continue](https://routerhub.si/guides/vscode-continue)
- [VS Code with Cline](https://routerhub.si/guides/vscode-cline)
- [Claude Code](https://routerhub.si/guides/claude-code)
- [Codex](https://routerhub.si/guides/codex)
- [Claude Desktop](https://routerhub.si/guides/claude-desktop)

## Machine-readable references

- Catalogue: https://routerhub.si/models.json
- OpenAPI: https://routerhub.si/openapi.json
- Pricing and feature notes: https://routerhub.si/models