Replicate API Tester
Build, run and copy Replicate prediction API requests as cURL, Python and JavaScript.
Building the Replicate request and the cURL, Python and JavaScript code happens entirely in your browser. When you press Run, your API token and prompt are sent directly from your browser to Replicate — never to ArrayKit. Your token is kept in browser memory only and is never stored or sent to ArrayKit. Avoid using production tokens on shared machines.
Try the OpenAI API Tester
About Replicate API Tester
The Replicate API tester lets you build, run and copy a Replicate predictions request without leaving your browser. Enter a model as owner/name, type an input prompt and paste your Replicate API token, and ArrayKit assembles the POST to /v1/models/{owner}/{name}/predictions with the Prefer: wait header for you. The request and the ready-to-paste cURL, Python and JavaScript snippets are generated locally on your device. When you press Run, the call goes directly from your browser to Replicate so you can see the real prediction output, then copy whichever code form you need into your project. It is built for developers, ML engineers and prompt testers who want to try a model, confirm a token works or grab a working snippet fast — no Postman collection or local SDK setup required.
Features
- Targets the Replicate predictions endpoint: POST /v1/models/{owner}/{name}/predictions
- Specify any public model by owner/name with built-in suggestions like meta/meta-llama-3-8b-instruct
- Generates copy-paste cURL, Python (requests) and JavaScript (fetch) snippets
- Adds the Prefer: wait header so you get the finished prediction in one call
- Bearer token authentication using your Replicate API token (r8_…)
- Press Run to send the request directly to Replicate and view the live response
- Token is held in browser memory only — never stored or sent to ArrayKit
- Request building and code generation happen entirely on your device
How to use the Replicate API Tester
- Paste your Replicate API token (starts with r8_) into the token field
- Enter the model as owner/name, e.g. meta/meta-llama-3-8b-instruct
- Type your input prompt in the prompt box
- Press Run to send the prediction request and read the response
- Switch the cURL, Python or JavaScript tab and copy the snippet
Example
Input
curl -X POST 'https://api.replicate.com/v1/models/meta/meta-llama-3-8b-instruct/predictions' \
-H 'Authorization: Bearer r8_xxx' \
-H 'Content-Type: application/json' \
-H 'Prefer: wait' \
-d '{"input":{"prompt":"Write a haiku about TypeScript."}}'
Output
{ "status": "succeeded", "output": ["Strict types align..."] }
A model prediction request built by the tester, ready to run from a terminal.
Common errors & troubleshooting
- 401 Unauthorized when you press Run. — Your token is wrong or missing. Paste a valid Replicate API token from replicate.com/account/api-tokens — it should start with r8_.
- 404 Not Found for the model. — Check the model is written as owner/name (e.g. meta/meta-llama-3-8b-instruct) and that the model exists and is accessible to your account.
- 422 Unprocessable Entity about the input. — The model expects different input fields. Confirm the model's input schema on its Replicate page; this tester sends input.prompt by default.
- Request fails or hangs in the browser (CORS / network). — Copy the generated cURL and run it from your terminal, or use the Python/JavaScript snippet server-side where browser CORS rules do not apply.
- Response shows status starting/processing instead of output. — The Prefer: wait header is included to return a finished prediction, but long jobs can still time out — re-run or poll the prediction URL from your own code.
Frequently asked questions
- What is the Replicate API tester?
- It is a browser tool that builds a Replicate predictions request from a model (owner/name), an input prompt and your API token, then runs it and gives you copy-ready cURL, Python and JavaScript code.
- Which Replicate endpoint does it call?
- It POSTs to https://api.replicate.com/v1/models/{owner}/{name}/predictions with a Bearer token and the Prefer: wait header, sending {"input":{"prompt":"..."}} as the body.
- Is my Replicate API token safe?
- The request and snippets are built locally, and your token is kept in browser memory only — it is never stored or sent to ArrayKit. When you press Run it goes directly from your browser to Replicate. Avoid using production tokens on shared machines.
- Can I test any Replicate model?
- Yes — enter any model you can access as owner/name. The tester sends an input.prompt field by default, so models that take a different input schema may need you to adjust the generated code.
- Why run the cURL in a terminal instead of the browser?
- Most Replicate calls work from the browser, but if a request is blocked by CORS or your network, the generated cURL or server-side Python/JavaScript snippet will run cleanly outside the browser.
- Do I need to install the Replicate SDK?
- No. The tester produces plain cURL, requests-based Python and fetch-based JavaScript, so you can try a model and copy a working call without any SDK setup.
Related tools
- OpenAI API Tester — Build, run and copy OpenAI Chat Completions API requests as cURL, Python and JavaScript.
- Anthropic Claude API Tester — Build, run and copy Anthropic Claude Messages API requests as cURL, Python and JavaScript.
- Google Gemini API Tester — Build, run and copy Google Gemini generateContent API requests as cURL, Python and JavaScript.
- Hugging Face API Tester — Build, run and copy Hugging Face Inference API requests as cURL, Python and JavaScript.
- API Request Client — Send HTTP requests, build headers and params, inspect responses — your last 25 requests are saved locally.
- cURL to Fetch — Convert a curl command to a JavaScript fetch() call.
- URL Parser — Break a URL into its parts and list query parameters.
- Basic Auth Generator — Generate and decode HTTP Basic Auth Authorization headers from a username and password.
All ArrayKit tools