ModelHubby scout badgeModelHubbyFIELD MANUAL · EST. 2026

← All guides · provider record →

Requesty — free tier setup

What you get free

Requesty is a router/aggregator — one OpenAI-compatible key in front of many upstream providers, with automatic fallback, caching, and load balancing. Two free things stack: signup credits and a free-models allowance. Catalog raw_text: "Official docs: "New accounts include free credits to start routing immediately." (amount unstated). Free-LLM: "Standard | Free monthly credits | Free monthly credits | Free tier included | Multi-provider routing"." The live pricing page adds a $0 plan with "200 requests per day on free models", no credit card, no platform fee. Category: trial_credit — but note the flagged possibility that credits renew monthly (Free-LLM claim, unconfirmed officially; if it verifies, this recategorizes toward permanent_free). The gotcha: the free credit amount is unpublished — check your dashboard balance before relying on it.

Friction check
Credit card no for the free tier ("No credit card required", pricing page); unknown for signup credits generally (catalog: unknown)
Phone verification unknown
Org/membership no
Est. time to first key ~3–5 min
Trains on your prompts no — "Requesty is a pass-through gateway. We don't train on your requests or responses." (catalog, from official docs); prompt logging: unknown

Get a key

  1. Sign up at https://app.requesty.ai/sign-up (confirmed, quickstart).
  2. Go to the API Keys page: https://app.requesty.ai/api-keys (confirmed, quickstart).
  3. Create a key (exact button name not detailed in the docs — unconfirmed, verify on first signup) and copy it.
  4. Check your credit balance in the dashboard — the free-credit amount is unpublished (location of balance display unconfirmed — verify on first signup).

First request

Base URL from the catalog (OpenAI-compatible): https://router.requesty.ai/v1. An Anthropic-compatible surface also exists at https://router.requesty.ai/anthropic/v1 (catalog). Model ids are provider/model form; catalog's confirmed example is openai/gpt-4o (note: this is a paid upstream model — it draws from your credit pool, not the 200/day free-models allowance; free-model ids were not confirmed, browse the model list in the app).

export REQUESTY_API_KEY="YOUR_KEY"
curl https://router.requesty.ai/v1/chat/completions \
  -H "Authorization: Bearer $REQUESTY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "openai/gpt-4o", "messages": [{"role": "user", "content": "Say hi in one word"}], "max_tokens": 16}'

Expected success shape: OpenAI-style chat.completion JSON with choices[0].message.content.

Limits that will bite you

  • Free credit amount is unpublished (catalog: "amount unstated") — you cannot budget against it blind. Monthly renewal is claimed by Free-LLM but unconfirmed; assume one-time signup credits until your balance visibly refills.
  • 200 requests/day on free models on the $0 plan (pricing page) — a hard daily count, and it applies only to free models, not the full 400+ catalog.
  • Shared credit pool across all routed models (catalog: shared_pool true) — one chatty upstream drains credits for everything.
  • Aggregator semantics: upstream models are third-party; "OpenAI-compatible" at the router does not guarantee identical behavior (tool calling: unknown in catalog).
  • Streaming usage accounting: to get token usage on streamed responses you must send stream_options: {"include_usage": true} (catalog notes) — without it your local spend tracking undercounts.
  • Paid path is pay-as-you-go with a 5% markup on base model costs (pricing page) — cheap, but it means exhausted credits fail over to billing only if you've added payment, otherwise requests fail.

Key rotation runbook

  • Second key: create another key from https://app.requesty.ai/api-keys (multiple keys assumed supported; unconfirmed — verify on first signup).
  • Revoke: delete the key on the same API Keys page (unconfirmed — verify on first signup).
  • ModelHubby integration: export REQUESTY_API_KEY_2="..." — the key pool picks it up automatically; state machine cools down rate-limited keys per-key. Note: keys under one account share the same credit pool and 200/day free-model counter, so rotation only helps if the limits are per-key (unconfirmed).

ModelHubby usage

pnpm modelhubby show requesty
pnpm modelhubby probe --live --provider requesty --apply