Keeping accounts connected

Every connected account syncs through its own proxy, around the clock. When the proxy stops working, background syncing for that account stops with it. This page covers what the errors mean, what happens automatically, and the routes and events you can build on.

What happens when a proxy fails

1. One failure never alerts you. A request that fails because of the proxy marks the account proxy_error and schedules a re-test of the saved proxy from our server about 15 minutes later.
2. A confirmed failure notifies you once. If that re-test also fails, we emit account.proxy_failed and keep re-testing on a growing schedule: 15, 30, 60, 120 and 240 minutes, then at most every 6 hours.
3. Recovery is automatic. The first re-test that passes clears the error, resumes background syncing and emits account.proxy_recovered. Saving a new proxy with PATCH /accounts/{of_user_id}/proxy schedules a re-test within about a minute.

While an account is in proxy_error, its background jobs (polling, subscriber, transaction and campaign refreshes) are held, so a dead proxy does not burn through your proxy plan. A login rejected by the platform is a different state (login_failed); fixing the proxy does not clear it.

What each error means

The code on connection_error (and on the events below) tells you what went wrong. The codes are the same ones the API has always returned.

Check health and data use

curl "$BASE/api/crm/$CRM/accounts/$OFUID/proxy/health" -H "X-API-Key: $KEY"

{
  "success": true,
  "of_user_id": "482687148",
  "state": "failing",
  "code": "proxy_auth",
  "title": "Your proxy provider refused the login (407)",
  "message": "This usually means the proxy plan is out of data, has expired, or the proxy password was changed.",
  "advice": "Check your proxy dashboard, then press Test.",
  "failed_at": "2026-10-15T02:53:20",
  "next_retest_at": "2026-10-15T03:53:20",
  "background_paused": true,
  "last_test": { "at": "2026-10-15T03:23:20", "ok": false, "code": "proxy_auth", "latency_ms": null, "of_reachable": null },
  "usage": {
    "today":   { "bytes_up": 1048576, "bytes_down": 9437184, "requests": 214 },
    "last_7d": { "bytes_up": 9961472, "bytes_down": 88080384, "requests": 2140 },
    "daily": [{ "day": "2026-10-15", "bytes_up": 1048576, "bytes_down": 9437184, "requests": 214 }]
  },
  "recommendation": { "kind": "static_isp_unlimited", "text": "…" }
}

state is ok, failing, untested or no_proxy. The proxy itself is never returned. Data figures are counted from request and response sizes plus an estimate of connection overhead, so your provider's figure may be slightly higher.

The account list (GET /accounts) carries the same summary per account: proxy_state, proxy_failed_at, proxy_next_retest_at, proxy_bytes_today, proxy_bytes_7d, and connection_error.next_retest_at, auto_retesting and background_paused.

For every account at once:

curl "$BASE/api/crm/$CRM/proxy/usage?days=7" -H "X-API-Key: $KEY"

Test the saved proxy

curl -X POST "$BASE/api/crm/$CRM/accounts/$OFUID/proxy/test" -H "X-API-Key: $KEY"

{ "success": true, "ok": true, "code": null, "title": null, "message": null,
  "latency_ms": 231, "of_reachable": true, "resumed": true }

This tests the proxy already saved on the account, from our server, and sends nothing in the body. When the account was in a proxy error and the test passes, the error clears and syncing resumes straight away (resumed: true). Limited to 6 tests per minute. To test a proxy you haven't saved yet, use POST /proxy/test with {"proxy": "…"} as described in Proxies.

Events

{
  "of_user_id": "482687148",
  "username": "creatorname",
  "platform": "onlyfans",
  "code": "proxy_auth",
  "title": "Your proxy provider refused the login (407)",
  "message": "This usually means the proxy plan is out of data, has expired, or the proxy password was changed.",
  "failed_at": "2026-10-15T02:53:20",
  "recovered_at": null,
  "downtime_seconds": null,
  "retest_count": 1,
  "link": "/dashboard/accounts/482687148"
}

On account.proxy_recovered, recovered_at and downtime_seconds are filled in. Subscribe with "event_types": ["account.proxy_failed", "account.proxy_recovered"] on a webhook, or * for everything.

Choosing a proxy that stays up

Most failures we see are pay-per-GB plans running out of data, and rotating proxies changing IP mid-session.

• Unlimited data. Static ISP and dedicated proxies are usually sold per IP with unlimited data. A per-GB residential plan runs dry, and then every request gets a 407. Use the usage figures above to size a plan if you must pay per GB.
• Sticky sessions. If your provider rotates IPs, use a sticky (session) port that holds one IP for as long as possible, ideally 24 hours or more. An IP that changes on every request looks like a new login each time.
• One proxy per account, in the creator's own country, and keep the same region when you replace it.