HTTP Error Simulator — Test Error Handling with Any Status Code Online

Constructor

Every field is optional. The request method is irrelevant and the body is ignored — all parameters travel in the query string.

Status code(s)One code (100-599) or a comma-separated list a random code is drawn from, e.g. 500 or 500,503,429. Default 500.Error rateProbability the error fires, 0-1. Below 1, the remaining requests answer a clean 200 {"ok":true}. Default 1.Delay (ms)Fixed or random delay before answering, e.g. 300 or 100-800. Maximum 30000. Default 0.SequencePer-key cycling list, e.g. 200,200,503,429: the n-th request answers the n-th code, then it repeats. Takes priority over status and rate. Requires a key.Scenario keyScenario id the sequence counter is stored under (letters, digits, _ . : -; max 128 chars). Required when a sequence is set.Provider presetBody format: Auto (preset default)Retry-After (s)Overrides the Retry-After header in seconds. A 429 always carries one.Body formatDrop probabilityProbability, 0-1, that the connection is aborted instead of answered. The client observes a network error.SeedDeterministic 32-bit RNG seed so the status/delay sequence is reproducible.HangSend the headers and keep the connection open until the delay (or 30s) elapses.

Live URL preview

Rebuilt as you edit. This is the exact URL you call.

https://arttron.dev/api/fail?status=500&rate=1

Response

Fill in the constructor and send a request to see the status, headers, body and elapsed time.

One-click examples

Call it from your code

curl

curl -i 'https://arttron.dev/api/fail?status=500&rate=1'

fetch

const res = await fetch('https://arttron.dev/api/fail?status=500&rate=1');
console.log(res.status, await res.text());

axios

import axios from 'axios';
try {
  const res = await axios.get('https://arttron.dev/api/fail?status=500&rate=1');
  console.log(res.status, res.data);
} catch (err) {
  console.log(err.response?.status, err.response?.data);
}

What is an HTTP error simulator?

An HTTP error simulator answers a request with a deliberately wrong response so you can exercise the failure path of a client. Instead of unplugging a service or waiting for a real outage, you ask for an exact status — say 503, or 429 — and the tool returns it. Point your code at the simulator URL and watch how it behaves. This page is the browser front end for the /api/fail endpoint: the same engine that powers the simulate_http_error MCP tool.

Realistic failures are rarely a lone 500. A flaky API fails only some of the time, a rate limit hands back a Retry-After, a provider stamps its own error shape (OpenAI, Stripe, AWS XML, Cloudflare), a gateway stalls before timing out and a connection can simply drop. The simulator reproduces each of those: an error rate from 0 to 1, a fixed or random delay, a per-key sequence for retry tests, a Retry-After override, provider-shaped bodies in JSON, text, HTML or XML, a dropped connection and a hung response. A seed makes the random status and delay reproducible.

How to use the HTTP error simulator

Set the status (or a comma-separated list a random one is drawn from) and, if you want intermittent failures, an error rate below 1.
Add a delay, a sequence with a scenario key, a Retry-After override, a provider preset and a body format as needed.
Copy the live URL into curl, fetch or axios — or press Send to call it right here and read the status, headers, body and elapsed time.
Switch to one of the one-click examples (flaky API, rate limit, retry, slow, drop) to see a realistic failure in one move.

Frequently asked questions about the HTTP error simulator

Why is a dropped connection not a real disconnect?

The Next.js App Router exposes no raw socket API, so a true socket destroy is unavailable. The route sends the response head and then aborts the body stream, so clients observe a network error (terminated / fetch failed) rather than a clean status. The response carries X-Sim-Drop: true so the emulation is explicit.

How does the sequence mode work?

With seq and a key, the n-th request for that key answers the n-th code in the list, then the sequence repeats. The counter lives per key for one hour and is refreshed on every hit, so a retry test keeps its order across a whole run.

Is this tool free and does it need a key?

The site front end calls /api/fail on the site plane, so no key is needed from this page. Scripts may authenticate with an API key; the endpoint is rate limited to 60 requests per minute per IP.

Back to home