TAOTOK
5-minute quickstart

Make your first API call in 5 minutes

One API key. One OpenAI-compatible endpoint. GPT-4o, Claude, Gemini, DeepSeek and Kimi K2 — switch models by changing a single string.

Do you already have a Taotok API key?

1

Get your API key

skip if you already have one

Pick an option above and this step adapts. Or read both:

Already have a key

  • Export it in your shell: export TAOTOK_KEY="sk-…"
  • Never commit it. Use an env var or a secret manager.
Open dashboard

Need a key

  1. Sign up with your email — about 30 seconds.
  2. Go to API Keys → Create new key.
  3. Copy it once and save it in a password manager.
Create account & key
2

Pick a model and copy the snippet

Choose a model and a language, then hit Copy. Everything is pre-filled — the only thing you supply is your key.

Base URL https://api.taotok.io/v1

Model IDs are case-sensitive and must match your dashboard's Models list exactly. Not sure? Run: curl https://api.taotok.io/v1/models -H "Authorization: Bearer $TAOTOK_KEY"

3

Run it and read the response

Paste it into your terminal, or run the script. A 200 with text inside choices[0].message.content means you are live.

✓

You're live. Three things worth doing next

Most people stop at the first successful call. These three steps are what turn a first call into a habit.

Swap the model

Same code, different string. Try DeepSeek V3 for the cheapest tokens, Claude for long documents, Kimi K2 for agentic runs.

Compare models & prices →

Watch your usage

Open Usage to see tokens, spend and remaining balance. Set a monthly budget so a run never surprises you.

Open usage →

Join the Discord

Ask about models, report a failing endpoint, or request a rate-limit bump. Real devs answer fast.

Join Discord →
?

If the call fails

Almost every first-call failure is one of these four. Work top to bottom — about two minutes.

401 — invalid api keyauth
  • Echo the variable: echo $TAOTOK_KEY. Empty means your shell never exported it.
  • Check the header literally: Authorization: Bearer YOUR_KEY — one space after Bearer, no quotes around the key.
  • Re-copy the key. A trailing space or line break pasted from the dashboard is the #1 cause.
  • Confirm the base URL is https://api.taotok.io/v1 — not http, no double /v1/v1.
  • Still stuck? Disable the key in the dashboard and issue a fresh one, then retry immediately.
429 — rate limitedlimits
  • Read the Retry-After response header and sleep that long, then retry.
  • Add exponential backoff: 1s, 2s, 4s, 8s with a bit of jitter.
  • Lower max_tokens and shorten the prompt — you are billed on tokens, and limits follow tokens.
  • Check your RPM/TPM tier on the Usage page; request a bump in Discord with your use case.
  • Bursty first tests (loops of 50 calls) trip limits that real traffic never will. Test with 1 call first.
404 / model not found — or a weird model namemodel
  • List what you can actually call: curl https://api.taotok.io/v1/models -H "Authorization: Bearer $TAOTOK_KEY"
  • IDs are case-sensitive. gpt-4o and GPT-4O are not the same thing.
  • Use the gateway ID, not the vendor's marketing name ("DeepSeek V3" → deepseek-chat).
  • Some models are plan-gated. If it is missing from /v1/models, ask in Discord to have it enabled.
  • Sanity check in the browser first: run the same prompt in the Playground to isolate gateway issues from your code.
402 / insufficient balance or quotabilling
  • A first test call costs a fraction of a cent — if it fails on balance, the account simply has zero credit.
  • Open Usage to confirm the remaining balance before topping up.
  • Top up in USDT. TRC-20 is usually the cheapest and fastest to confirm.
  • Wait for the required confirmations; credit lands after the network confirms, not the moment you send.
  • Set a low-balance alert so this never stops a production job.
Timeout, TLS error, or nothing comes backnetwork
  • Use https://. Plain http is not served.
  • Behind a corporate proxy or VPN? Retry from a different network to rule it out.
  • Add -v to curl to see the real status code and headers.
  • Confirm the path is /v1/chat/completions — a missing /v1 returns HTML or 404.
  • Cross-check in the Playground. Works there but not in your code? It is your request body, not the gateway.

Keep building — you've done the hard part

The first call is the whole battle. From here it is just swapping model strings and watching spend.

See pricing & models Ask a question on Discord