🎁 AgentRouter — $175 in free AI API credits: every referral link lives here

Give-Away and Freebies
, ,

AI-generated summary

:bullseye: What you get: Up to $175 in free API credits for multiple AI models via the AgentRouter referral train.

:white_check_mark: Start here:

  • AgentRouter — log in via GitHub or LinuxDO to claim credits.
  • API Key generation — create a key under Tokens and claim daily check-ins.

:toolbox: Also useful:

  • OpenAI-compatible clients — plug the custom base URL into Codex or similar apps.
  • Local proxy setup — fixes occasional 401 unauthorized client errors.

:warning: Watch out:

  • Credits cannot be withdrawn, brand-new GitHub accounts are rejected, and rates vary by model.
1

Every link feeds the next — AgentRouter free credits
Every link feeds the next — AgentRouter free credits2400×1120 191 KB

AgentRouter — $175 free, one key, every model

:brain: One thread, every link. Join once, grab your own referral link, post it here — someone joins through it and you’re both paid.

What lands on day one :money_mouth_face:

:wrapped_gift: what arrives :dollar_banknote: amount
welcome credit $100
:dollar_banknote: referral bonus $50
:date: first check-in $25
= your starting stack $175
:repeat_button: every day after, one tap +$25

That is ≈ $750 a month of API credits for a daily click — a frontier-model bill the program pays, not your card. (The check-in rate is theirs; confirm it on your dashboard.)

Where your own link comes from :backhand_index_pointing_down:

  1. :key: Sign in with GitHub — no email, no card.
  2. :compass: Open the invite page on your dashboard — your personal link is sitting there.
  3. :clipboard: Copy it and reply here with that one line. That’s the whole ask.

Setup, in one drawer :gear: — key, Codex config, the 401 fix, batches:

⚙️ Setup — API key, Codex config, the 401 recipe, batches

Your key. Dashboard → Tokens → new key → tap check-in daily for the +$25. Balance refreshes after a logout/login.

Codex or any OpenAI-compatible client:

model = "gpt-6-astra"
model_provider = "agentrouter"

[model_providers.agentrouter]
name = "AgentRouter"
base_url = "https://co.agentrouter.org/v1"
env_key = "AGENTROUTER_API_KEY"
wire_api = "chat"

Swap gpt-6-astra for gpt-6.1-sol (unlimited, 24/7), claude-opus-5 or deepseek-v4-flash — whatever your dashboard lists.

The 401 fix. If a client throws 401 unauthorized client detected, a lightweight local proxy fixes it in two minutes — OpenCode’s own docs cover the setup.

Limited daily batches. Claude and GPT go out in daily batches at 02:00 and 11:00 UTC; the batch for that model has to be live.

Before you commit. The check-in refuses brand-new GitHub accounts and LinuxDO login works as the other door; there are no top-ups, credits are earned only; they pay API calls and never withdraw; unsupported clients bill at about double.

Amounts move. The welcome credit isn’t pinned (it has run $150–$200 across recent months); the numbers above are the current offer — confirm on the signup page.

:speech_balloon: Post your link below — one line, no pitch, and it’s on the shelf.

2

:wrapped_gift: Start here — my link: https://agentrouter.org/register?aff=QfPV

Sign up through it and the $175 lands on your account automatically. Then grab your own link and post it right below — the next person joins through yours.

3

Starting with one small task is the safest way to judge a router like this, because a short prompt shows latency and cost before any real work depends on it. I would check two things first: whether the credit has an expiry date, and whether each model keeps its own rate limit behind the shared key. Keeping a direct provider key as a fallback also helps, so a router outage does not stop a job halfway through. If the first test feels slow or inconsistent, that is a good signal to pause rather than move a bigger project over. A quick note of which model handled which task well makes the next choice much easier.