# Sandbox (test keys)

> Noptcha test keys (nop_test_) return fixed answers without charges; magic values simulate failures, timeouts, proxy errors and slow tasks.

Source: https://noptcha.com/docs/sandbox/

Tasks created with a **test key** (`nop_test_…`) never reach a solver, are never charged and never appear in usage. They behave like real tasks otherwise: same endpoints, same validation, same response shapes. Use them to build and test integrations, and to run the examples in this documentation.

## Answers

| Type | Sandbox solution |
| --- | --- |
| Token types (reCAPTCHA, hCaptcha, Turnstile) | `token` = `NOPTCHA_TEST_TOKEN_<task id>`, `user_agent` = `Mozilla/5.0 (Noptcha test)` |
| `funcaptcha` | only `token` |
| `cloudflare.challenge` | `cookies.cf_clearance` = `NOPTCHA_TEST_TOKEN_<task id>`, `user_agent` |
| `image.text` | `text` = `noptcha` |
| Image grids and `image.funcaptcha` | `objects` = `[0]` |

Each [task type page](https://noptcha.com/docs/task-types.md) shows the exact sandbox answer in both APIs. The compatibility API returns the same answer in its own field names (for example `gRecaptchaResponse`).

## Simulating outcomes

Put a magic value in `website_key` (compatibility API: `websiteKey`), `website_public_key` (`websitePublicKey`) or `question`:

| Value starts with | Outcome |
| --- | --- |
| `test-unsolvable` | fails with `unsolvable` (`ERROR_CAPTCHA_UNSOLVABLE`) |
| `test-timeout` | fails with `timeout` (`ERROR_CAPTCHA_UNSOLVABLE`) |
| `test-proxy-error` | fails with `proxy_error` (`ERROR_PROXY_CONNECT_REFUSED`) |
| `test-slow-<seconds>` | stays `processing` for that many seconds (at most 90), then succeeds |

In the sandbox the task is always created, and a failure is reported when you read the task: `getTaskResult`, `res.php`, `GET /v1/tasks/{id}`, or the response of `POST /v1/tasks` when it waits. Live tasks of some types can also fail in the create response itself (see [Solving a task](https://noptcha.com/docs/compatibility-api.md#solving-a-task)); handle both. A `test-slow-5` task is a good way to exercise polling and `Prefer: wait`.

## Limits

The sandbox has its own limits per organization, on top of the per-key limits ([Rate limits](https://noptcha.com/docs/limits.md)):

|  | Limit |
| --- | --- |
| Tasks created | 10 per second (`429 rate_limited` / `ERROR_NO_SLOT_AVAILABLE`) |
| Tasks still processing (`test-slow-…`) | 20 (`429 concurrency_limited`) |
| Synchronous waits | 20 per API server; more return `202` at once |

These are starting values and may change.

## What else is different

-   Sandbox tasks carry `"sandbox": true` in v1; the compatibility API returns an empty `ip`.
-   Results are kept for 10 minutes, like live tasks.
-   Test keys see only sandbox tasks and live keys only live tasks.
-   `GET /v1/usage` with a test key returns an empty report with `"sandbox": true`.
-   Balance endpoints return your organization’s real balance, which test tasks never change.