# Mingles Router > One OpenAI-compatible inference API. Point your existing OpenAI client at `https://router.mingles.ai/v1`, keep your code, pay 6–35× less than Claude and GPT APIs (blended 4:1 input/output; verified November 9, 2026). Default model `deepseek-ai/DeepSeek-V4-Flash-0731`. Free API key in 30 seconds, no card, no KYC. Pay by card or crypto (USDT/USDC). Mingles Router is an LLM gateway that speaks the OpenAI dialect — `/v1/chat/completions`, `/v1/models`, streaming, and tool calling — so any OpenAI SDK or agent framework (Cursor, Cline, OpenClaw, Hermes, OpenHands, Kilo Code, Aider, Continue, n8n, LangChain, LlamaIndex) works after a single `base_url` change. Requests are served by open-weight models on a decentralised inference network, with automatic fallback to the next model when one is unavailable. Three values are all any integration needs: - base URL: `https://router.mingles.ai/v1` - API key: created at https://router.mingles.ai/app - model: `deepseek-ai/DeepSeek-V4-Flash-0731`, `MiniMaxAI/MiniMax-M2.7`, `zai-org/GLM-5.3-Flash` Full text of every documentation page in one file: https://router.mingles.ai/llms-full.txt ## Docs - [Quickstart](https://router.mingles.ai/docs): The three values — base URL, key, model name — and worked examples in curl, Python and JS. - [Model limits & behaviour](https://router.mingles.ai/docs/limits): Context windows, max output tokens, rate limits, per-model notes. - [Tools & function calling](https://router.mingles.ai/docs/tools): OpenAI-compatible `tools` / `tool_calls`, including raw-passthrough mode. - [Request headers](https://router.mingles.ai/docs/headers): Supported and passthrough headers, including `X-Tool-Mode`. ## API - [API base URL](https://router.mingles.ai/v1): OpenAI-compatible endpoint (`/v1/chat/completions`, `/v1/models`). - [OpenAPI spec](https://router.mingles.ai/openapi.json): Machine-readable JSON spec. - [Status](https://router.mingles.ai/status): Live service status. ## Models - `deepseek-ai/DeepSeek-V4-Flash-0731` (default) — DeepSeek-V4-Flash (default, cheapest, tool-calling) - `MiniMaxAI/MiniMax-M2.7` — MiniMax-M2.7 (tool-calling) - `zai-org/GLM-5.3-Flash` — GLM-5.3-Flash ## Pricing - [Pricing](https://router.mingles.ai/pricing): Plans (Mini $9, Standard $29, Team $79/seat) and per-token rates; card via Stripe or USDT/USDC. - [Routing modes](https://router.mingles.ai/routing): Route each request by cost and data sensitivity. ## Agent integrations - [Agent hub](https://router.mingles.ai/agents): Set-up config for any OpenAI-compatible client. - [Hermes](https://router.mingles.ai/agents/hermes): NousResearch’s personal agent — native custom-endpoint wizard. - [OpenClaw](https://router.mingles.ai/agents/openclaw): Open-source personal agent with native custom-provider config. - [OpenAI Codex](https://router.mingles.ai/agents/codex): OpenAI’s terminal coding agent, pointed at our models instead of theirs. - [DeepSeek Harness](https://router.mingles.ai/agents/deepseek-harness): DeepSeek’s open-source coding agent — the Claude Code idea, running on your key. - [OpenHands](https://router.mingles.ai/agents/openhands): Autonomous coding agent (OpenAI-Compatible via LiteLLM). - [Kilo Code](https://router.mingles.ai/agents/kilo-code): VS Code / JetBrains coding agent — OpenAI-Compatible provider. - [Cursor](https://router.mingles.ai/agents/cursor): Override the OpenAI Base URL — bring your own key and model. - [Cline](https://router.mingles.ai/agents/cline): VS Code agent — “OpenAI Compatible” provider, three fields. - [Roo Code](https://router.mingles.ai/agents/roo-code): The most active Cline fork — same OpenAI-Compatible provider. - [Aider](https://router.mingles.ai/agents/aider): CLI pair-programmer — two env vars and a model flag. - [Continue](https://router.mingles.ai/agents/continue): VS Code / JetBrains chat + edit — one models block in config. - [n8n](https://router.mingles.ai/agents/n8n): One OpenAI credential with a Base URL override powers every AI node. - [Open WebUI](https://router.mingles.ai/agents/open-webui): Self-hosted ChatGPT-style UI — one OpenAI API connection. - [SillyTavern](https://router.mingles.ai/agents/sillytavern): Custom Chat Completion source — endpoint, key, model id. - [LangChain](https://router.mingles.ai/agents/langchain): ChatOpenAI takes a base_url — the rest of your chain is untouched. - [LlamaIndex](https://router.mingles.ai/agents/llamaindex): OpenAILike, with is_chat_model=True — the default is False and it will bite you. - [Vercel AI SDK](https://router.mingles.ai/agents/vercel-ai-sdk): createOpenAICompatible — a provider in four lines. - [LiteLLM](https://router.mingles.ai/agents/litellm): One openai/ prefix and an api_base — as a library or as a proxy. - [PydanticAI](https://router.mingles.ai/agents/pydantic-ai): OpenAIChatModel with an OpenAIProvider — typed agents, one base_url. - [CrewAI](https://router.mingles.ai/agents/crewai): LLM(custom_openai=True, base_url=...) — one object, shared by every agent. - [OpenAI Agents SDK](https://router.mingles.ai/agents/openai-agents-sdk): Works against any OpenAI-compatible endpoint — after two settings most people miss. - [LibreChat](https://router.mingles.ai/agents/librechat): One custom endpoint block in librechat.yaml. - [Zed](https://router.mingles.ai/agents/zed): An openai_compatible block in settings.json — max_tokens is required. - [opencode](https://router.mingles.ai/agents/opencode): A provider block in opencode.json, on the AI SDK openai-compatible package. ## Comparisons - [vs OpenAI API](https://router.mingles.ai/compare/openai): Drop-in replacement for `api.openai.com/v1`. - [vs OpenRouter](https://router.mingles.ai/compare/openrouter): Where each one fits. - [vs Claude API](https://router.mingles.ai/compare/claude-api): Published benchmark scores and prices, side by side, with sources. - [Cheapest API for Cursor](https://router.mingles.ai/answers/cheapest-api-for-cursor): Direct answer with a dated price table and Cursor setup steps. - [DeepSeek V4 Flash API](https://router.mingles.ai/models/deepseek-v4-flash): Pricing, published benchmarks and code samples for deepseek-ai/DeepSeek-V4-Flash-0731. Cheapest model we serve; plan sizes are quoted in it. - [MiniMax M2.7 API](https://router.mingles.ai/models/minimax-m2.7): Pricing, published benchmarks and code samples for MiniMaxAI/MiniMax-M2.7. - [Kimi K2.6 API](https://router.mingles.ai/models/kimi-k2.6): RETIRED — moonshotai/Kimi-K2.6 is no longer served; page kept for reference. - [GLM-5.2 API](https://router.mingles.ai/models/glm-5.2): RETIRED — zai-org/GLM-5.2-FP8 is no longer served (superseded by zai-org/GLM-5.3-Flash); page kept for reference. - [Models index](https://router.mingles.ai/models): The models we serve, with per-token prices and how fast each spends a plan allowance. ## Tools - [Token cost calculator](https://router.mingles.ai/tools/token-cost-calculator): Cost your monthly input/output tokens across every model in the dated price table. ## Answers - [Answers index](https://router.mingles.ai/answers): Direct answers about LLM APIs, pricing and setup. - [What is an OpenAI-compatible API?](https://router.mingles.ai/answers/openai-compatible-api): What the convention covers, what it does not, and how to verify a provider. - [Switching from the OpenAI API](https://router.mingles.ai/answers/switch-from-openai-api): The two lines that change and the four things that break. - [Free LLM API without a credit card](https://router.mingles.ai/answers/free-llm-api-no-credit-card): What a no-card free tier includes and what it does not. - [How many tokens does $1 buy?](https://router.mingles.ai/answers/tokens-per-dollar): Tokens per dollar across models at a 4:1 input/output ratio, with the arithmetic. - [Why is the API returning 429?](https://router.mingles.ai/answers/429-rate-limit): 429 vs 402 vs 503, and the retry pattern. - [Which model for a coding agent?](https://router.mingles.ai/answers/best-model-for-coding-agents): What agent loops spend tokens on and how to pick per step. - [vs Claude & ChatGPT subscriptions](https://router.mingles.ai/compare/claude-chatgpt): Token volume for the same monthly spend. ## For teams - [For agencies](https://router.mingles.ai/agencies): Cost routing for automation and content agencies. - [Partner program](https://router.mingles.ai/partners): 20% lifetime revenue share. ## Blog - [We vibe-coded a farm of Telegram agents in 2–3 days — here are the guts](https://router.mingles.ai/blog/vibecoding-the-farm): A dev diary of a Telegram agent farm: cheap inference as layer zero, Hermes as the runtime, MCP for tool security — including the rakes we hit. - [One router, three agents: how our support, admin and marketing bots share one cheap inference endpoint](https://router.mingles.ai/blog/one-router-three-agents): Support, admin and marketing bots on one shared brain: an OpenAI-compatible router, a Hermes runtime and an MCP tool layer — the full architecture. ## Optional - [Terms of Service](https://router.mingles.ai/legal/terms) - [Privacy Policy](https://router.mingles.ai/legal/privacy) - [Acceptable Use Policy](https://router.mingles.ai/legal/aup) - [Refund Policy](https://router.mingles.ai/legal/refund)