AI Tools Tool updated 1 day ago

AI API Gateway Check

The AI API Gateway Check tests any OpenAI, Anthropic, Gemini or OpenAI-compatible API in your browser. Pick a preset, paste endpoint, model and key, hit Test API — get status, latency and full response. Extras: Get Models lists models, Balance checks token quota with automatic endpoint fallback, Cancel aborts hung requests, plus history, import/export and auto-fix for endpoint URLs. Keys never leave your control.

Always Free Easy to use Instant results Private and secure

Use AI API Gateway Check

API Compatibility
Model ID / Name
API Endpoint
API Key (optional)
Authorization: ---
Your API key is never saved on our servers. It is kept only in this browser, for request history and exported configs.
Headers
Prompt
Response 200 OK

                
History
0
No requests yet

Read Content

Learn how the tool works and when to use it.

What Is This Tool

The AI API Gateway Check is a free online tester for AI chat APIs. It sends a real request to any OpenAI, Anthropic, Gemini or OpenAI-compatible endpoint and shows exactly what comes back: status code, response time and full response body. No signup, no install, everything runs in your browser.

Who Is It For

  • Developers integrating an AI API who want to verify keys, endpoints and models before writing code.
  • Self-hosters running local gateways (Ollama-style, custom proxies) who need a quick connectivity and auth check.
  • Resellers and aggregator users juggling third-party compatible endpoints who must confirm which base URL, model name and key actually work.
  • Non-technical users who received an endpoint + key from a provider and want to see if it is alive without touching a terminal.

What It Does

You describe the request (provider preset, endpoint, model, key, headers, prompt) and the tool builds the correct HTTP call for that provider's dialect, sends it, and renders the result. Around that core it adds model listing, balance checks, request history and config import/export, so a working setup can be saved and replayed later.

How to Use It

  1. Pick a preset matching your provider (or Custom).
  2. Paste the endpoint URL. Leave auto format on and the tool corrects partial hosts and wrong paths, offering a Use this fix when needed.
  3. Enter the model ID (or leave the preset default). Paste button included.
  4. Paste the API key if your endpoint needs one. Many local models work without any key; an upstream 401 tells you a real one is missing. Check the masked Authorization preview to confirm where the key will go.
  5. Type a prompt (default is ping) and click Test API.
  6. Read the response box: status badge, elapsed milliseconds, formatted body. Copy it or adjust and retry.

Presets Explained

  • OpenAI - Official /v1/chat/completions API. Example models: gpt-4, gpt-3.5-turbo.
  • OpenAI Compatible - Any provider speaking the OpenAI chat format (DeepSeek, Groq, Together and similar). Endpoint must end in /v1/chat/completions.
  • OpenAI URL Compatible - Providers with OpenAI format at a custom base path ending in /chat/completions. Common with free shared proxies.
  • Anthropic - Messages API (/v1/messages). Key travels as x-api-key with an anthropic-version header. Browsers block direct calls, so it always routes through the secure proxy.
  • Google Gemini - generateContent endpoint with the model name inside the URL. Key travels as a ?key= query parameter. Also proxied.
  • Custom - Anything else accepting JSON POST. Sends {"model": ..., "prompt": ...}.

Endpoint Auto-Format: What You Need to Know

You do not need a perfect URL. Paste a bare host, an IP, or a URL missing the API path and auto-format rebuilds it to the preset's canonical form. If it cannot match, it tells you the expected shape instead of failing silently. Turning the toggle off gives strict validation with no rewriting.

API Key & Security: What You Need to Know

The key field is optional by design. The preview line under it shows the exact header or query parameter that will carry the key, masked for safety. Keys live only in your browser (form + history + exported files) and are never stored on the server. Proxy on means the key passes through the server to reach the provider and is then discarded; proxy off means it goes straight from your browser to the provider. Share exported configs carefully: they contain the key.

Secure Proxy & SSL: What You Need to Know

Some providers block browser requests (CORS). The secure proxy toggle routes the call through the server instead, which is why Anthropic and Gemini work here. It only allows public IPs and refuses private-network targets. If you test a local HTTPS server with a self-signed certificate, verification fails on purpose; switch Verify SSL off (there is a one-click hint when this happens) and retry. Never disable it for production endpoints.

Headers: What You Need to Know

Most users never touch headers: Content-Type: application/json plus auth are set automatically. Add rows only when your provider demands extras (e.g. a custom version or routing header). Dangerous hop-by-hop headers (host, connection, cookies, forwarding headers) are stripped for you.

Test, Models, Balance, Cancel

  • Test API - The main check. Sends your prompt, shows status, latency and body. Use ping for the cheapest smoke test.
  • Get Models - Lists available models. The tool derives the /models URL from your endpoint and queries it through the proxy, so it works even on CORS-blocked providers.
  • Balance - Shows token quota (/api/usage/token/ on the endpoint host). Only credit-based compatible proxies answer it; official OpenAI, Anthropic and Gemini endpoints return 404, which is normal and not a bug.
  • Cancel - Appears while a request is in flight. Slow or hung upstreams can be aborted instead of waiting for the 60-second timeout. Cancelled runs are labelled Cancelled and never pollute history.

Reading the Response

A green 2xx badge means the provider accepted the call; 4xx usually means wrong key, model name or request shape; 5xx or proxy errors point at the provider being down. The millisecond timer helps compare providers. JSON is pretty-printed and highlighted; anything else shows as plain text. Copy grabs the raw text for tickets, docs or debugging.

History: What You Need to Know

Every Test API run is remembered locally (last 50, newest first) with preset, endpoint, status and timing - keys included, since it is your own browser. Click any entry to reload it into the form exactly as it ran. Delete one entry or clear everything when rotating keys or switching machines.

Import & Export: What You Need to Know

Export Config downloads the current setup as a JSON file you can keep, share or back up. Import Config loads such a file back; if it came from elsewhere and lacks a preset, the tool infers one from the API shape and host and tells you to double-check it. Sample downloads a minimal starter file showing the expected format.

Common Failures and Fixes

  • CORS blocked - Enable the secure proxy and retry.
  • 401 Unauthorized - Key missing or wrong; check the Authorization preview and preset.
  • 404 - Wrong endpoint path or model name; accept the suggested URL fix.
  • SSL / certificate error - Self-signed local cert; disable Verify SSL for local testing only.
  • Timeout - Provider slow or unreachable; Cancel and retry, or test with ping.

Comments (0)

For user discussion, edge cases, and follow-up tips.
Leave a comment
Your comment will appear publicly after submission.
No comments yet. Be the first to comment!
Let Us Know!
Report an issue with this tool

Describe the issue you encountered so we can investigate and improve the tool.

What helps most
Include the input, expected result, actual result, browser/device, and a screenshot when helpful.
The screenshot button captures the page in the browser and attaches the generated image to this form.
Full-page screenshot
Captured in your browser with newisty.
No screenshot captured yet.
Tool Report