FAQ

Account & keys

Q: I created a key but can't find it now.
A: Go to /console/keys — all your tokens are listed there. Copy it from the list. If you need to roll it, create a new one and delete the old one from the same page.

Q: Do Claude Code and Codex need separate API keys?
A: No. One APIMaster key works in both. Claude Code uses ANTHROPIC_AUTH_TOKEN (or ANTHROPIC_API_KEY on older versions); Codex stores it in ~/.codex/auth.json as OPENAI_API_KEY. The key format is the same — only the config file and variable name differ.

Q: The console panel is blank or shows a white screen after I sign in.
A: Sign out and sign back in at /login. The console session can expire after a server restart — this is the only fix.

Pricing

Q: How much does Claude API actually cost through APIMaster?
A: APIMaster channels run about 85% below Anthropic list price. Official Anthropic rates are: Haiku 4.5 at $1/$5, Sonnet 4.6 at $3/$15, Opus 4.8 at $5/$25 (input/output per 1M tokens). APIMaster's effective rate for Sonnet 4.6 is around $0.45/$2.25. Check the marketplace for live per-channel prices — they vary and update in real time.

Q: What payment methods does APIMaster accept?
A: epay, PayPal, credit card, and USDT. Minimum top-up is $1. There is no monthly subscription or seat fee — you pay only for tokens used.

Q: What is the cheapest model for high-volume tasks?
A: Claude Haiku 4.5 at ~$0.15/$0.75 (APIMaster rate) or DeepSeek V4 Flash at ~$0.10/$0.20 are the cheapest capable options. For most coding tasks Haiku 4.5 is good enough; for multi-step reasoning use Sonnet 4.6. See cheapest LLM API for a full comparison.

Q: Is there a free tier?
A: No. The minimum top-up is $1 with no expiry. You are charged per token with no base fee.

Setup — Claude Code

Q: How do I connect Claude Code to APIMaster?
A: Edit ~/.claude/settings.json (create it if missing):

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://apimaster.ai",
    "ANTHROPIC_AUTH_TOKEN": "YOUR_APIMASTER_KEY"
  }
}

https://apimaster.ai is the base host — no /v1. If your Claude Code version uses ANTHROPIC_API_KEY instead of ANTHROPIC_AUTH_TOKEN, use that. Then run claude and pick a model with /model claude-sonnet-4-6. Full guide: Claude Code setup.

Q: Claude Code keeps asking me to log in to Anthropic. How do I stop it?
A: ANTHROPIC_BASE_URL is either missing or set to the wrong value. It must be exactly https://apimaster.ai — not https://apimaster.ai/v1, not https://apimaster.ai/v1. Open ~/.claude/settings.json and check the value. Also confirm the file is in the right folder (%userprofile%\.claude on Windows, ~/.claude on macOS/Linux).

Setup — Codex

Q: How do I connect Codex to APIMaster?
A: Create two files in ~/.codex/:

config.toml:

model_provider = "apimaster"
model = "gpt-5.4"

[model_providers.apimaster]
name = "apimaster"
base_url = "https://apimaster.ai/v1"
wire_api = "responses"
requires_openai_auth = true

auth.json:

{ "OPENAI_API_KEY": "YOUR_APIMASTER_KEY" }

Then run codex. Full guide: Codex setup.

Setup — other tools

Q: How do I connect Dify / Open WebUI / LiteLLM to APIMaster?
A: All three use an OpenAI-compatible base URL. Set it to https://apimaster.ai/v1 and paste your APIMaster key. For Dify: Settings → Model Provider → add custom OpenAI provider. For Open WebUI: Settings → Connections → add OpenAI API connection. For LiteLLM: set api_base=https://apimaster.ai/v1 in your model list config. Specific guides: Dify, Open WebUI, LiteLLM.

Q: How do I connect Cursor to APIMaster?
A: In Cursor → Settings → Models → add a custom OpenAI-compatible provider with base URL https://apimaster.ai/v1 and your key. Pick any model id from supported models. See Cursor guide.

Troubleshooting

Q: I get model not found. What's wrong?
A: The model id is misspelled or not available on your account. Check the exact id at supported models — copy-paste it, don't retype. Also confirm your account has balance.

Q: Should base_url include /v1 or not?
A: Yes for all OpenAI-compatible tools (Codex, Cline, Cursor, Open WebUI, Dify, LiteLLM): use https://apimaster.ai/v1. No for Claude Code: use https://apimaster.ai. Mixing them up is the most common setup mistake.

Q: I get 401 Unauthorized even with the right key.
A: Check for invisible characters — paste the key into a plain text editor first, copy it again, and paste into config. Also confirm the account has a positive balance at /console/dashboard.

Q: What is the difference between /v1/messages and /v1/chat/completions?
A: /v1/messages is the Anthropic protocol used by Claude Code and the anthropic Python SDK. /v1/chat/completions is the OpenAI protocol used by everything else. APIMaster supports both. If you see a route not found error, your client is sending one protocol to the wrong endpoint — check which base URL your tool expects.

Model verification

Q: How do I know if a channel is actually serving the model it claims?
A: Use the API key tester — paste any endpoint and key and it runs a fingerprint probe. Results show a TrustScore and the most likely model family. You can also check the marketplace for channel detection history — channels with repeated weekly passes are the safer choice.

Q: What is a TrustScore?
A: A probability estimate that a channel is serving the claimed model, based on behavioral fingerprinting. A high score (e.g. 90%+) means the output patterns closely match that model family. It is a risk signal, not a legal proof — a score of 85% means 15% uncertainty, not that 15% of requests are wrong.

Q: Can a provider fake a model name?
A: Yes — any provider can label a route with any model name. That is why APIMaster runs weekly fingerprint checks instead of just trusting the API's model field. Before routing production traffic to an unverified channel, run a few test prompts through the tester.

Security

Q: Do you store my API key when I use the tester?
A: No. The fingerprint tester uses your key to make test requests and does not persist it. Do not submit keys with billing admin permissions — use a scoped key with only inference access.

Q: What should I watch out for with cheap API providers?
A: The three common risks: (1) routing to a weaker model than advertised, which degrades output quality silently; (2) usage logs that undercount or overcount tokens, breaking your budget; (3) rate limits that only show up at scale. Check detection history in the marketplace before committing to a channel.

Comparisons

Q: How does APIMaster compare with OpenRouter?
A: OpenRouter has more model coverage (200+ vs ~50 major providers at APIMaster). APIMaster's differentiator is fingerprint verification — weekly behavioral checks on every listed channel — plus USDT and epay payment options that OpenRouter doesn't offer. If you need a broad model catalog, OpenRouter is wider; if you need model authenticity signals and flexible payment, APIMaster is the better fit.

Q: How does APIMaster compare with the official Anthropic API?
A: The official Anthropic API gives guaranteed model authenticity, direct support, and priority access to new features. APIMaster is roughly 85% cheaper for Claude, requires no Anthropic account, and accepts more payment methods, but adds one relay hop and relies on fingerprinting rather than official guarantees. For regulated or compliance-sensitive production, prefer the official API. For cost-sensitive development, batch jobs, or teams without an Anthropic account, APIMaster is a practical alternative.

Q: OpenRouter already has a free tier. Why pay for APIMaster?
A: OpenRouter's free tier has low rate limits and does not include model verification. APIMaster starts at $1 with real token allocations and every channel has a public detection history. If you are running anything beyond quick tests, the verification layer and predictable pricing are worth the minimum top-up.

Billing

Q: Which price is the final price I actually pay?
A: Marketplace prices are indicative. Your actual charge is in the console at /console/log — that is the authoritative record. If marketplace price and invoice differ, the invoice is correct.

---

Still stuck? Go to Contact and include your error message, the tool you're using, and what base_url you configured.