One API for
every AI model.
Image, video, voice and LLMs. Corent picks the model, reroutes on failure, and bills only finished results.
- 1000+Models
- 50+Labs
- 4Modalities
- 1API key
Type anything.
Watch it decide.
Watching it decide is free.
Anyone can see the decision, the price and the code, no sign-in needed. Sign in to generate free on Air and Lite, with a daily limit. Premium and above run on credits, from $5, on the same one-request API.
- 01ProfileIt reads the prompt.
- 02ScoreIt ranks every live model.
- 03RouteIt sends the work to the best one.
- 04RerouteOne fails. It re-picks in 2ms.
- 05DeliverOne result, with the exact cost.
Every model, behind one call.
- 00:00.0 route opened
- 00:14.2 provider_error
- 00:14.2 rerouted in 2ms
- 00:23.7 delivered
the failed attempt, on your bill: $0.00
Every response
carries a receipt.
"meta": { "model": "image-air", "cost_cents": 3 }See full pricing →Four prompts. Four finished films.
Every one made through the same call, on the same balance. Touch one and it comes into colour.

PlateI
“a colossal chrome starship igniting and lifting off a desert launchpad at night, volumetric light beams, cinematic slow motion”

PlateII
“high-production anime action: a young adventurer in a red coat leaping through a pirate ship's rigging over a sunlit sea”

PlateIII
“a polished porcelain emblem set into black embossed paper, hexagonal relief tiling, one raking studio light”

PlateIV
“a lit sigil monolith standing on a cracked alien plain at dusk, twin moons low over distant spires”
1 key.
- Nano BananaGoogle
- FLUXBlack Forest Labs
- SeedreamByteDance
- GPT Image 2OpenAI
- Ideogram V3Ideogram
- Qwen ImageAlibaba
- Recraft V4Recraft
- MAI ImageMicrosoft
- HiDreamHiDream
- Stable DiffusionStability
- KreaKrea
- ImagenGoogle
- Veo 3.1Google
- KlingKuaishou
- SeedanceByteDance
- WanAlibaba
- HailuoMiniMax
- PixVersePixVerse
- LTX-2Lightricks
- Gen-4Runway
- RayLuma
- Hunyuan VideoTencent
- SoraOpenAI
- MochiGenmo
- Eleven v3ElevenLabs
- Gemini TTSGoogle
- GPT-5OpenAI
- GeminiGoogle
- KimiMoonshot
- DeepSeekDeepSeek
- QwenAlibaba
- LlamaMeta
- MistralMistral
- GrokxAI
- CommandCohere
- NovaAmazon
You describe it.
The machines make it.
Set a ceiling.
Agents never spend past it.
"corent": { "command": "npx", "args": ["corent-mcp"] }
"max_cost_cents": 50