Run thousands of open-source AI models (image gen, LLMs, speech, video) via Replicate, paid per-call through MPP/Tempo.
What it does
This endpoint provides MPP-gated access to Replicate's vast catalog of open-source AI models. By sending a POST request with a model owner, model name, and input parameters, you can run inference on models spanning image generation (e.g. Flux Schnell, Stable Diffusion), language models, speech recognition, video generation, and more. Payment is handled via the Locus MPP protocol using the Tempo settlement method, with per-call charges that vary by model (described as $0.005–$0.05 per run).
The service exposes four endpoints under the same MPP gateway: `/replicate/run` to execute a model, `/replicate/get-prediction` to retrieve the status/output of a previous run, `/replicate/get-model` to fetch metadata about a specific model, and `/replicate/list-models` to paginate through available models. The auxiliary endpoints have fixed costs (e.g. $0.003 for get-prediction, $0.004 for get-model and list-models, assuming the currency uses 6 decimals). The run endpoint's cost is model-dependent and not fixed in the challenge.
The request schema for the run endpoint requires `model_owner` and `model_name` as strings, plus an `input` field whose schema varies per model — consult the model's page on replicate.com for specifics. No API key is needed; authentication and billing are handled entirely through the MPP payment flow on Tempo L2.
Capabilities
Use cases
- —Generate images from text prompts using models like Flux Schnell or Stable Diffusion without managing API keys
- —Run language model inference on open-source LLMs with micropayment-based billing
- —Transcribe audio or generate speech using Replicate-hosted speech models
- —Build AI agent workflows that dynamically select and run models from a large catalog
- —Retrieve model metadata and paginate available models to discover new capabilities
Fit
Best for
- —AI agents needing on-demand access to diverse open-source models without subscription commitments
- —Developers who want to pay per inference call via crypto micropayments rather than managing API keys
- —Prototyping and experimentation across many model types (image, text, audio, video) through a single gateway
Not for
- —High-volume production workloads where direct Replicate API access with bulk pricing would be cheaper
- —Use cases requiring guaranteed sub-second latency, as the MPP payment handshake adds overhead
- —Users who need fine-tuning or model training — this endpoint only supports inference
Quick start
curl -X POST https://replicate.mpp.paywithlocus.com/replicate/run \
-H "Content-Type: application/json" \
-d '{
"model_owner": "black-forest-labs",
"model_name": "flux-schnell",
"input": { "prompt": "a cat sitting on a rainbow" }
}'Example
Request
{
"input": {
"prompt": "a cat sitting on a rainbow"
},
"model_name": "flux-schnell",
"model_owner": "black-forest-labs"
}Endpoint
Quality
The OpenAPI spec is well-structured with clear request schemas, pricing hints, and multiple endpoints documented. However, the probe did not capture a live 402 challenge (the endpoint is POST-only and was probed with HEAD/GET), no response examples are available, and the crawl pages all returned 404. Pricing details rely on the x-payment-info description field rather than a confirmed challenge.
Warnings
- —Probe returned 404 on HEAD/GET — endpoint is POST-only; liveness not confirmed via 402 challenge
- —No example response schema or sample output available in the spec
- —Currency address 0x20c000000000000000000000b9537d11c60e8b50 is not a standard well-known token; decimal count assumed to be 6 based on the $0.005–$0.05 description
- —Input field schema varies per model and is not documented in this spec — consult replicate.com for model-specific parameters
Citations
- —Run thousands of open-source AI models via API — image generation, language models, speech recognition, video, and more.https://replicate.mpp.paywithlocus.com
- —Model-dependent pricing of $0.005–$0.05 per runhttps://replicate.mpp.paywithlocus.com
- —API reference available at replicate.com/docs/reference/httphttps://replicate.com/docs/reference/http
- —LLM-friendly docs at beta.paywithlocus.com/mpp/replicate.mdhttps://beta.paywithlocus.com/mpp/replicate.md