Overview
Welcome to the Concurred API documentation
What is Concurred?
Concurred provides a single API to access 9 top AI models plus media generation (speech, music, video). Think of it as a universal AI gateway — one API key, any model.
Base URL
https://agent-heavy.vercel.app
Two APIs, One Key
| Gateway API | Chat API | |
|---|---|---|
| Endpoint | /api/v1/chat/completions | /api/chat |
| Format | OpenAI-compatible (drop-in) | Custom SSE / JSON |
| Use case | Direct model access, production integrations | Multi-model battles, debates, voting |
| Modes | Single model | Chat, Battle, Fight, Multi-instance |
| Web search | No | Yes (Battle, Fight) |
| Voting & leaderboard | No | Yes (Fight mode) |
| Fallback / caching / guardrails | Yes | No |
| BYOK | Yes | No |
Use the Gateway API when you want a drop-in OpenAI replacement with advanced gateway features. Use the Chat API when you want models to compete, debate, and vote.
Key Features
- 9 AI Models — Claude, GPT, Grok, Gemini, DeepSeek, Kimi, Mistral, Llama, MiniMax. See Models for full details.
- Gateway Features — Fallback routing, response caching, load balancing, guardrails, prompt templates, BYOK.
- Media Generation — Speech (TTS), music, and video via MiniMax. See Media API.
- Fashion AI — Virtual try-on, model creation, background editing. See Fashion API.
- Multi-Model Vision — Image analysis across 4 models. Non-vision models get AI-generated descriptions automatically.
- Live Web Search — In battle and fight modes, models autonomously search the web with citations.
- Observability — Full request tracing, cost tracking, and usage analytics.
Authentication
Create an API key in the dashboard. Keys start with ck_. See Authentication for details.