OpenRouter vs Claude Code setup · As of 2026-11-19

OpenRouter Alternative for Claude Code and Codex CLI: Price the Fees First, Then Decide

As of November 19, 2026, if all you want is one key for many models, OpenRouter already does that: its FAQ states plainly that inference is passed through at the underlying provider's list price with no markup. The 5.5% fee ($0.80 minimum) is charged when you buy credits, 8% on Business, plus 5% for crypto payments. What actually pushes developers to look for an OpenRouter alternative is three things — top-up fees, payment rails, and client protocol coverage: Standard and Business list cards, AliPay, and USDC, while Claude Code speaks Anthropic Messages and Codex CLI speaks Responses. Check those three points before moving, and if you do move, QCode is one per-token gateway among several, with rates on /models and no price comparison from us.

Updated 2026-11-19

Switch models on the same endpoint and measure the cost per token yourself; check /models for callable models.

#5.5% credit fee#3 env vars#Two protocol endpoints#Payment self-check

Four numbers that show exactly where OpenRouter charges you

5.5%

Platform fee when buying credits (Standard)

Official wording: OpenRouter charges a 5.5% ($0.80 minimum) fee when you purchase credits. Business is 8%, and inference itself is passed through at the underlying provider's list price with no markup, so the extra line appears only at top-up time.

5%

Crypto payments carry another 5%

The FAQ states crypto payments carry a separate 5% fee. If you habitually top up with USDC or another coin, budget that 5% on top of the credit-purchase fee; they are two separate charges, not one.

$25,000

Monthly BYOK list-price allowance

When you bring your own provider keys, Standard and Business include $25,000 of list-price inference per month with no BYOK fee; usage above that costs 5% of what the same model and provider would normally cost on OpenRouter, deducted from your credits.

3 variables

The complete Claude Code to OpenRouter config

ANTHROPIC_BASE_URL goes to https://openrouter.ai/api, your OpenRouter key goes into ANTHROPIC_AUTH_TOKEN, and ANTHROPIC_API_KEY must be explicitly blank — the official guide marks that line Important and explains why under Why ANTHROPIC_AUTH_TOKEN.

What this is: the exact step where an OpenRouter alternative becomes attractive

OpenRouter is a gateway that routes many models behind one key. Its official Claude Code integration guide states that it exposes an input compatible with the Anthropic Messages API, so pointing ANTHROPIC_BASE_URL at it lets Claude Code speak its native protocol with no local proxy server, with usage (including reasoning tokens) billed against your OpenRouter credits and shown in its dashboard. People searching for an OpenRouter alternative usually are not short on models — the pricing page lists 600+ models and 80+ providers on Standard and Business — they are stuck on three concrete frictions.

  1. 1A fixed cost at the top-up step: every credit purchase carries 5.5% with a $0.80 minimum, which weighs more on small top-ups, and crypto adds 5%.
  2. 2Payment rails: Standard and Business list credit and debit cards, AliPay, and crypto (USDC); if you need WeChat Pay or multi-chain USDT / USDC / BTC / ETH, you have to look elsewhere.
  3. 3Protocol coverage: Claude Code speaks Anthropic Messages, Codex CLI speaks Responses, and whether one gateway can feed both clients matters more day to day than catalogue size. Tick those three boxes against your own setup and the decision follows.

November 18–19, 2026: a wave of comparison posts, and the community doing its own math

As of 2026-11-19

  1. 1On 2026-11-18 a cluster of English-language OpenRouter alternative write-ups appeared, including a byoklist guide that names several gateways claiming no markup — a community roundup, not a vendor announcement.
  2. 2On 2026-11-19 a developer post did the arithmetic out loud: The fee is 5.5%, or $0.80 on a $10 top-up, alongside One account and one API key give me pay-as-you-go — which says the demand is a single metered key, not another monthly plan. Those community figures were computed against one person's top-up amount; they are not a quote from OpenRouter and your own bill follows your own purchase.

Three dates: the comparison wave, the community math, and the check date

2026-11-18

A cluster of English-language OpenRouter alternative write-ups lands, including byoklist's openrouter-alternatives guide naming several gateways that claim no markup; this is a community roundup, and OpenRouter itself has published no list of substitutes.

2026-11-19

A developer post puts a number on the fee: 5.5% works out to $0.80 on a $10 top-up, and notes that one account and one API key give pay-as-you-go access; community discussion around that time also shows demand for one key across several models.

2026-11-19

Check date: the OpenRouter FAQ, pricing page, and Claude Code integration guide were captured as published text, and QCode model IDs were taken from the qcode.cc/models list of the same day, with endpoint URLs and variable names verified word for word.

Confirmed official fees and setup vs community claims and unannounced details

Confirmed (OpenRouter documentation, captured 2026-11-19)

  1. 1Where the fee sits: a 5.5% fee with a $0.80 minimum applies when you buy credits, with Standard at 5.5% and Business at 8%; that platform fee is tied to credit purchases rather than individual requests.
  2. 2Inference has no markup: pricing is passed through at the underlying provider's list price.
  3. 3Crypto: crypto payments carry a separate 5% fee.
  4. 4BYOK: Standard and Business include $25,000 of list-price inference per month with no BYOK fee, then a 5% fee above that, deducted from credits.
  5. 5Claude Code setup: the base URL is https://openrouter.ai/api, the key goes in ANTHROPIC_AUTH_TOKEN, ANTHROPIC_API_KEY is explicitly blank, a cached Anthropic login is cleared with /logout, and /status confirms which variable supplies the auth token.
  6. 6Failed attempts: an errored attempt is not charged for its model tokens, so with fallback you only pay for the run that answers.

Community claims, and what has not been published

  1. 1Community only: the 2026-11-18 write-ups name several gateways that claim no markup, and the 2026-11-19 developer post converts 5.5% into $0.80 on a $10 top-up. Those numbers assume a particular top-up amount; someone else's arithmetic is not your bill.
  2. 2Not published: OpenRouter's docs give no per-request latency or reliability figures, so measure it yourself in the same client if it matters.
  3. 3Not published: QCode's own rates always point to /models; endpoints, protocols, and payment rails are the verifiable details.
  4. 4Compliance boundary: standard API key configuration for local CLI tools is the supported workflow — no workarounds for regional restrictions, rate-limit evasion, or third-party identity verification.

Stay on OpenRouter vs move to a per-token gateway

Stay on OpenRouter: a broad catalogue, with the fee concentrated at top-up

Per the official pricing page, Standard and Business both list 600+ models and 80+ providers, inference passes through at the provider's list price, and the platform fee is charged only on credit purchases at 5.5% (8% on Business); failed and fallback attempts are not billed for model tokens. The trade-off is that the fee is tied to top-ups: the $0.80 minimum bites harder on small purchases and crypto adds another 5%. If your card, AliPay, or USDC works and you do not need a second gateway, staying is a reasonable call.

Move: one key feeding Claude Code and Codex CLI separately

Switching gateways changes the billing entry point and the payment rails, not the models. On QCode, one cr_ key covers three setups: Claude Code and other Anthropic Messages clients use https://api.qcode.cc/api, Codex CLI uses base_url = "https://api.qcode.cc/openai" in config.toml with wire_api = "responses", and OpenAI-compatible clients use https://api.qcode.cc/openai/v1.

Billing is per token with rates on /models, and payment options include Alipay, WeChat Pay, Visa / Mastercard, Apple Pay, Google Pay, and USDT / USDC / BTC / ETH. We do not compare the two providers' prices; run the checklist below against your own usage.

Do this: OpenRouter's official three lines, and what changes on QCode

  1. 1OpenRouter's documented setup: export ANTHROPIC_BASE_URL="https://openrouter.ai/api"; export ANTHROPIC_AUTH_TOKEN="$OPENROUTER_API_KEY"; export ANTHROPIC_API_KEY="" — the guide labels that third line Important: Explicitly blank out ANTHROPIC_API_KEY to prevent conflicts.
  2. 2On QCode only the first two lines change: ANTHROPIC_BASE_URL=https://api.qcode.cc/api and ANTHROPIC_AUTH_TOKEN=your cr_ key, with ANTHROPIC_API_KEY left empty again, otherwise the old key is sent as x-api-key and shadows the gateway credential; then source your shell profile or open a new terminal.
  3. 3Verify before you work: run /status in Claude Code and check that the Auth token line names ANTHROPIC_AUTH_TOKEN and the base URL is your endpoint, running /logout first if you ever signed in with an Anthropic account; for Codex CLI set wire_api = "responses" and base_url = "https://api.qcode.cc/openai" (no /v1 suffix) in ~/.codex/config.toml, and point OpenAI-compatible clients at https://api.qcode.cc/openai/v1.

Where QCode fits: one cr_ key, three setups, rates only on /models

QCode.cc is an independent API access service with no affiliation to Anthropic, OpenAI, or other model vendors. One cr_ key covers three setups: Claude Code and Anthropic Messages clients use https://api.qcode.cc/api, Codex CLI uses base_url = "https://api.qcode.cc/openai" with wire_api = "responses" in ~/.codex/config.toml, and OpenAI-compatible clients use https://api.qcode.cc/openai/v1.

Billing is per token and every live rate is on /models. Callable model IDs include claude-fable-5-2, claude-opus-5-5, claude-sonnet-5-5, gpt-6.2-sol, gpt-6.2-astra, kimi-k3.1-pro, deepseek-v4.1-flash, glm-5.3, and qwen3.8-max. Payment options: Alipay, WeChat Pay, Visa / Mastercard, Apple Pay, Google Pay, and USDT / USDC / BTC / ETH. Not ready for a monthly plan?

Pay-as-you-go Packs start at $4 for a one-time quota you use within 7 days; if you already have a Personal Plan, add a booster pack when you need more quota.

Frequently asked questions

Which lines do I actually change to point Claude Code at OpenRouter, and why must ANTHROPIC_API_KEY be empty?

Three variables, and the third must be an empty string. The official guide sets ANTHROPIC_BASE_URL="https://openrouter.ai/api", ANTHROPIC_AUTH_TOKEN to your OpenRouter key, and ANTHROPIC_API_KEY="". Its explanation is that ANTHROPIC_API_KEY is instead sent as x-api-key , is treated as a direct-Anthropic credential, and prompts you once in interactive mode to approve it over a saved login. Claude Code sends ANTHROPIC_AUTH_TOKEN as an Authorization Bearer token, while ANTHROPIC_API_KEY would be treated as a direct-Anthropic credential, so leaving it set makes Claude Code fall back to Anthropic auth and shadow the gateway token.

OpenRouter says there is no markup, so why is my balance short after a top-up?

Because no markup applies to inference, and the fee is charged when you buy credits. The FAQ says inference is passed through at the underlying provider list price with no markup, while a fee applies to credit purchases: 5.5% with a $0.80 minimum on Standard, 8% on Business, plus 5% for crypto payments. Per-request pricing therefore matches provider list price, and the difference appears at the moment of purchase.

What does one key running both Claude Code and Codex cost per month?

There is no fixed number — it follows token volume and the models you pick. On OpenRouter it is list price plus the credit-purchase fee; on QCode it is per-token billing with every rate listed on /models. No single monthly total applies because the bill depends on context length, cache hits, and concurrency. The practical approach is to run a week of real tasks on a small balance, then reconcile both platforms' per-request records instead of trusting someone else's monthly figure.

I have no international card, only Alipay, WeChat, or USDT. Does switching gateways fix that?

It depends which rail is missing. OpenRouter's pricing page lists credit and debit cards, AliPay, and crypto (USDC) for Standard and Business; if what you need is WeChat Pay or multi-chain stablecoins, QCode lists Alipay, WeChat Pay, Visa / Mastercard, Apple Pay, Google Pay, and USDT / USDC / BTC / ETH. Both accept Alipay and crypto, so the difference sits in WeChat and card rails — check against the one you can actually use.

Do I have to rebuy models or switch editors? And what is the $25,000 BYOK allowance?

No new client is needed; only the base URL and key change. Claude Code stays Claude Code and Codex CLI stays Codex CLI, with edits limited to environment variables and base_url in ~/.codex/config.toml. The $25,000 is OpenRouter's BYOK allowance: with your own provider keys, Standard and Business include $25,000 of list-price inference per month with no BYOK fee, then a 5% fee above that, deducted from credits. It waives OpenRouter's BYOK charge, not the provider's own bill.

How do I check whether a gateway's no-markup claim holds up?

Run four checks and ask for the source text each time. ① Where the charge sits: is it on top-up, per request, or a multiplier applied to the unit price? OpenRouter, for example, charges 5.5% ($0.80 minimum) on credit purchases and nothing on requests. ② Whether the payment rails fit you: cards, Alipay, WeChat, and USDT separately, not just one currency. ③ Protocol coverage: Claude Code needs an Anthropic Messages endpoint, Codex CLI needs Responses, OpenAI-compatible clients need /v1 — can they publish all three URLs? ④ Whether the bill reconciles per request: usage detail showing model and token counts per call is what makes verification possible. If any of the four comes only as a verbal promise, top up later.

Sources

OpenRouter official FAQ (https://openrouter.ai/docs/faq, captured 2026-11-19); OpenRouter official pricing (https://openrouter.ai/pricing, captured 2026-11-19); OpenRouter official Claude Code integration guide (https://openrouter.ai/docs/guides/coding-agents/claude-code-integration, captured 2026-11-19); community: byoklist's openrouter-alternatives guide (2026-11-18) and a developer post (2026-11-19); QCode: qcode.cc/models and docs.qcode.cc (checked 2026-11-19).

Get the three variables right, then choose where to keep your balance

Create a QCode account to get a cr_ key for Claude Code at https://api.qcode.cc/api and Codex CLI at https://api.qcode.cc/openai; if you would rather not commit monthly, start with a one-time Pay-as-you-go Pack, with per-model rates on /models.

Not sure where to start?

Try it for a week

Small Pack: $4 for $30 of quota, valid 7 days

Use it every day

Starter: $9.00/month with $5 of quota every day — just $0.060 per $1 of quota, vs about $0.113 for the Medium Pack