Replicate→
NinjaChatRun image, video, and language models through one API.
One key and balance for images, video, chat, search, and agent tools.
from ninjachat import NinjaChatimport osclient = NinjaChat(api_key=os.environ["NINJACHAT_API_KEY"])image = client.images.generate(model="flux-2-pro",prompt="Editorial product photograph",)169 text models39 image models12 video models21 MCP tools
One API for models, media, search, and tools.
NinjaChat API
GPT Image 2/v1/images/generations
See exactly where every request went.
Example tracesettledRequested modelclaude-sonnet-4.6
/v1/chat/completionsLatency842 msUsage1,238 tokensCost$0.00142
Set the spending limit before anything runs.
Control spend by project, API key, and agent.
ProjectProduction
enforcedSpent$184.20
Remaining$315.80
Monthly limit$500
$184.20 spent$315.80 remaining$500 limit
server API keykey$86.40 / $300enforceddesign agentagent
$72.80 / $120enforcedresearch agentagent
$25.00 / $80enforcedGive agents tools to finish the work.
Generate, edit, reuse, and save assets through NinjaChat MCP.
user request
Change the shoe upper to safety orange. Keep the sole, lighting, and framing unchanged. Save the final campaign image.
01Read sourcerunning
02Edit imagewaiting
03Check resultwaiting
04Write final assetwaiting
05Savedwaiting
Choose the gateway that fits your stack.
NinjaChatBest when you want:
- One account should cover media and frontier LLMs.
- You want a fixed price before each job.
- Apps and agents should share one model catalog.
ReplicateBest when you want:
- You need the largest community-model marketplace.
- Experimental model versions are the core job.
- Compute-time billing already fits your workflow.
Image models39 curated modelsThousands of community models
Video models12 curated modelsCommunity video catalog
Chat models169 production LLMsCommunity-hosted LLMs