# Vibetool > Vibetool is one OpenAI-compatible gateway for vibe coding tools — frontier AI models (chat, image, video) plus production-grade agent tools (search, translate, OCR, embedding, rerank), behind a single API. This file is a concise summary written for LLMs and AI agents that need to understand what Vibetool is, how to call it, and what guarantees it makes. For authoritative human-facing documentation, see . For the live model catalog, GET . ## Operator Vibetool is operated by **INGLITE INC.**, a technology company registered in Wyoming, USA. Contact: team@vibetool.ai. Founder: Alex Jiang (). ## What Vibetool is Vibetool is a unified neural gateway. One API key calls every major model and every common agent tool. The gateway routes requests across upstream providers and absorbs their differences (auth, schemas, rate limits, downtime). ## Trust commitments These are platform guarantees, not aspirations. Every claim below is enforced in code, not marketing copy: - **Zero data retention** — your prompts and responses are not used to train any upstream model. Vibetool does not log raw prompt/completion content beyond what billing requires. - **99.9% uptime, automatic failover** — when an upstream provider degrades, requests transparently fail over to the next provider in a strict priority order. - **No silent model downgrades** — the model id you request is the model id we send upstream. The mapping is one-to-one and stored in the database; we do not substitute Sonnet with Haiku or similar. - **Enterprise-sourced APIs** — every upstream is a B2B-licensed channel (no reverse-engineered keys, no individual trial accounts). ## API surface All endpoints are under the base URL **`https://api.vibetool.ai`** and authenticated with a Bearer token: ```http Authorization: Bearer YOUR_API_KEY ``` ### Chat completions (OpenAI-compatible) `POST /v1/chat/completions` — drop-in replacement for OpenAI's chat API. Supports streaming, function/tool calling, vision input, and thinking/reasoning models. Models include the Anthropic Claude, OpenAI GPT, Google Gemini, xAI Grok, DeepSeek, Moonshot Kimi, Alibaba Qwen, Zhipu GLM, and others. ### Image generation `POST /v1/images/generations` — submit an image generation task. Returns a `task_id`; poll `GET /v1/images/status/{task_id}` until status is `succeeded` (or `failed`). Models include Nano Banana, GPT-4o-image, Seedream, Z-Image, Wan, Qwen Image. ### Video generation `POST /v1/videos/generations` — submit a video generation task. Same async pattern: returns `task_id`, poll `GET /v1/videos/status/{task_id}`. Models include Sora, Veo, Kling, Seedance, Wan. ### Embeddings `POST /v1/embeddings` — text embeddings, OpenAI-schema-compatible. Models include OpenAI text-embedding-3-small/large, Jina embeddings v3/v4, Google Gemini embeddings, Zhipu embeddings. ### Tool endpoints `POST /v1/tools/{vendor}/{action}` — dedicated endpoints for tools whose schemas don't fit chat completions. Each has its own request/response shape; see the per-tool docs. - Search: `/v1/tools/perplexity/search`, `/v1/tools/exa/search`, `/v1/tools/exa/contents`, `/v1/tools/exa/answer`, `/v1/tools/bocha/web-search`, `/v1/tools/bocha/ai-search` - Translate: `/v1/tools/deepl/translate`, `/v1/tools/youdao/translate` - OCR: `/v1/tools/sophnet/ocr` - Rerank: `/v1/tools/jina/rerank` - Jina RAG suite: `/v1/tools/jina/embedding`, `/v1/tools/jina/segmenter`, etc. ### Listing the live catalog The authoritative list of currently-available models, with their canonical slug, modality, context length, and per-token pricing, is always: `GET /v1/models` (OpenAI-compatible) or `GET /api/models` (richer JSON with pricing fields). Do not hard-code model lists from this file — the catalog changes weekly. Always fetch live. ## Pricing model - Chat models: priced per 1M input tokens and per 1M output tokens (USD). Some models have tiered pricing (e.g. Gemini and Claude charge a higher rate above 200K context). - Image generation: priced per image (USD). - Video generation: priced per second or per video, depending on model. - Embeddings: priced per 1M tokens. - Tools: vary — some per request, some per 1M characters, some per 1M tokens. See `/pricing` or the per-tool docs page. There are no hidden fees, no markup that varies by request, and no discount tiers. The price you see in `/v1/models` is the price you pay. ## How to integrate 1. Sign up at and get an API key from the dashboard. 2. For chat completions, point your existing OpenAI SDK at `https://api.vibetool.ai/v1` with your Vibetool API key — that's it. 3. For image/video, use the async pattern (submit → poll task_id). 4. For tools, see the per-tool docs page on docs.vibetool.ai. ## Code example (Python, OpenAI SDK) ```python from openai import OpenAI client = OpenAI( api_key="YOUR_VIBETOOL_API_KEY", base_url="https://api.vibetool.ai/v1", ) resp = client.chat.completions.create( model="claude-sonnet-4-6", messages=[{"role": "user", "content": "Explain attention in one paragraph."}], ) print(resp.choices[0].message.content) ``` ## What Vibetool is good for - Drop-in replacement for OpenAI when you want access to non-OpenAI models (Claude, Gemini, Grok, etc.) without rewriting client code. - Production apps that need automatic failover when one upstream provider has an outage. - Multi-modal applications that mix chat, image, video, embedding and search behind one auth surface. - Teams that want a single billing and observability surface across all their AI usage. ## What Vibetool is not - Not a model trainer or fine-tuning platform. - Not a vector database. The embedding endpoints return vectors; storage/search is your responsibility. - Not a hosting layer for arbitrary user-uploaded models. ## Resources - Documentation: - Live model catalog: - Models + tools page: - Pricing: - About: - Extended LLM-friendly content: End of llms.txt