Claude Code doesn't care who issues your key. It reads two environment variables, sends requests to whatever host those variables point to, and bills you for whatever that host charges. That's the whole mechanism behind claude code api key pricing on a resale route: same CLI, same prompts, different bill.
This page shows the exact variables, the exact per-model rates, and the two endpoints that stop working if you switch. Prices checked March 2025.
Claude Code API Key Pricing: The Short Version
Token Master is a resale gateway. We buy upstream capacity and resell it through standard Anthropic-compatible endpoints. We are not Anthropic and we have no official partnership with Anthropic — your key just points at a different base URL.
Every model on this route is billed at 14.2% of its official list price. That's the entire pricing model. No tiers, no negotiated rates, no "contact sales." If a model costs $5.00 per million input tokens on Anthropic's list, it costs $0.71 here. That's claude api pricing with one number applied consistently, not a menu of plans.
The Claude Code Env Var That Controls Your Bill
Claude Code reads its credentials and endpoint from environment variables at startup. The two that matter for cost are the claude code api key itself and the claude code base url it gets sent to.
# Point Claude Code at a different billing endpoint
export ANTHROPIC_API_KEY="tm-xxxxxxxxxxxxxxxx"
export ANTHROPIC_BASE_URL="https://your-gateway-host/v1"
# Verify the CLI picked it up
claude --version
claude "say hello"Nothing else in your workflow changes. Same model names, same CLI flags, same tool-calling behavior. The only claude code env var work is setting the key and the base URL before you launch the CLI. See Anthropic's own reference on Claude Code settings and environment variables for the full list of variables Claude Code reads on startup.
What the Same Sessions Cost at 14.2% of List
Here is the per-model breakdown. "Our in / our out" is the price per million tokens on this route; "list in / list out" is Anthropic's published rate.
<div class="pricebox"><table><thead><tr><th>Model</th><th>Official list / 1M</th><th>Token Master / 1M</th><th>You pay</th></tr></thead><tbody><tr><td><code>claude-opus-4-8</code></td><td>$5.00 in / $25.00 out</td><td>$0.71 in / $3.55 out</td><td>14.2%</td></tr><tr><td><code>claude-sonnet-4-6</code></td><td>$3.00 in / $15.00 out</td><td>$0.43 in / $2.13 out</td><td>14.2%</td></tr><tr><td><code>claude-sonnet-5</code></td><td>$2.00 in / $10.00 out</td><td>$0.28 in / $1.42 out</td><td>14.2%</td></tr><tr><td><code>claude-haiku-4-5</code></td><td>$1.00 in / $5.00 out</td><td>$0.14 in / $0.71 out</td><td>14.2%</td></tr></tbody></table><p>Prices read from the live catalogue on 2026-10-22.</p></div>
| model | list in | list out | our in | our out | % of list |
|---|---|---|---|---|---|
| claude-opus-4-7-thinking | $7.50 | $52.50 | $1.06 | $7.46 | 14.2% |
| claude-opus-4-8-thinking | $7.50 | $52.50 | $1.06 | $7.46 | 14.2% |
| claude-opus-4-6-thinking | $7.50 | $42.50 | $1.06 | $6.04 | 14.2% |
| claude-opus-5-thinking | $7.50 | $37.38 | $1.06 | $5.31 | 14.2% |
| claude-opus-4-8 | $5.00 | $25.00 | $0.71 | $3.55 | 14.2% |
| claude-opus-4-7 | $5.00 | $25.00 | $0.71 | $3.55 | 14.2% |
| claude-opus-5 | $5.00 | $25.00 | $0.71 | $3.55 | 14.2% |
| claude-opus-4-6 | $5.00 | $25.00 | $0.71 | $3.55 | 14.2% |
Thinking variants bill at 1.5x the base model's upstream list price, at the same 14.2% percentage. That's why claude-opus-4-7-thinking costs more per output token than claude-opus-4-7 — the multiplier applies before the discount, not after.
Translated into actual Claude Code sessions, not just per-token rates:
| workload | official | Token Master |
|---|---|---|
| 500 chats, 4k input and 1.5k output each on claude-sonnet-4-6-thinking | $28.75 | $4.08 |
| 100 agent sessions, 50k input and 8k output each on claude-sonnet-4-6-thinking | $45.00 | $6.39 |
| batch classification, 40M input and 4M output on claude-haiku-4-5 | $60.00 | $8.52 |
That third row is listed for comparison only — batch classification at that scale needs the Batch API, which this route doesn't offer. More on that below.
Credit works through prepaid packs: $10 buys $50 of list-price usage credit, and that credit doesn't expire. Separately, Anthropic's own published rates are on their API pricing page if you want to check our numbers against list yourself. See current pack options on /#packages or run your own workload through the calculator.
Claude Code API Key vs Subscription
The claude code api key vs subscription question comes down to what you're paying for. A subscription bundles a fixed allotment of usage into a flat monthly fee. An API key bills per token, at whatever rate the key's issuer sets.
| Subscription | API key via Token Master | |
|---|---|---|
| Billing unit | Flat monthly fee | Per token, 14.2% of list |
| Usage cap | Fixed allotment | Pack credit, doesn't expire |
| Model access | Provider's bundled set | Models listed above |
| Batch/Files API | Varies by plan | Not offered |
| Overage | Often throttled | Buy another pack |
If your usage is spiky — heavy agent sessions one week, nothing the next — per-token billing on unused credit that doesn't expire tends to cost less than a flat subscription sized for your peak week. If your usage is steady and predictable, run the math with your own token counts before switching; the worked examples above are a starting point, not a guarantee for your workload.
Two Things That Break When You Migrate
This route covers chat, messages, streaming, tool calling, and vision. Two things Claude Code users sometimes rely on are not part of it:
Files and Batch endpoints. Not offered. If your workflow uploads files for reference or submits async batch jobs, those calls will fail against this base URL. The batch classification row in the table above is for price comparison only — you cannot actually run it here.
Org-level billing alerts. Anthropic's console-level spend alerts and invoicing are tied to an Anthropic organization account. Point Claude Code at a different base URL and those alerts stop seeing your usage, because the usage never touches Anthropic's billing system. You'll need to track spend through pack balance instead of org alerts.
Neither of these is a bug. They're the tradeoff for a route billed at 14.2% of list instead of Anthropic's published rate.
When Not to Use This
Don't use this route if:
- You need the Batch API or Files API for your workflow. They're not offered here, full stop.
- You depend on Anthropic's org-level billing alerts, invoicing, or SLA for compliance or finance reporting. Not offered.
- You need an official vendor relationship for procurement reasons. Token Master is a resale gateway, not an Anthropic partner, and we don't represent otherwise.
- Your usage is small enough that a $10 pack would sit mostly unused for months. Prepaid credit that doesn't expire is only an advantage if you'll eventually spend it.
If none of those apply, the per-token savings shown in the price table hold for any Claude Code session that uses chat, messages, streaming, tool calling, or vision.
Switch in Three Minutes
- Get a key from Token Master and a pack from /#packages.
- Set the two environment variables shown above, replacing the placeholder key.
- Run
claude "say hello"to confirm the CLI is hitting the new base URL. - Check your pack balance against actual Anthropic list pricing using the calculator before committing a full week's workload.
- Browse supported Claude models and routes in the docs or compare against Anthropic-routed pricing elsewhere on the site.
For a safe rollout, switch one non-production session first and keep the previous environment values in your shell history or secret manager. Confirm the returned model ID and token counts, then move the shared agent configuration. Reverting is only a base-URL and key change; no prompts, repositories, or Claude Code state need migration.
FAQ
Does Claude Code work with any API key, or only Anthropic's?
Claude Code reads its credentials and target host from environment variables at startup. It will send requests to whatever base URL you set, including a resale gateway, as long as the endpoint speaks the same API shape Claude Code expects.
What exactly is claude code api key pricing on this route?
Every model is billed at 14.2% of Anthropic's official list price, per token, with thinking variants billed at 1.5x the base model's list rate before that percentage is applied. There are no separate plan tiers.
Why is 14.2% different from the $0.20 credit price?
The $10-for-$50 pack is how you buy credit up front. The 14.2% figure is what Claude models specifically cost per token on this route, which is why the worked examples show savings bigger than a flat 20% discount would.
Can I still use prompt caching to cut costs further?
Prompt caching is an Anthropic-side mechanism for reducing repeated input token costs; see Anthropic's prompt caching docs for how it works. Whether it's available depends on the specific model and request shape, not on which key you use.
What happens to my pack credit if I stop using Claude Code?
It doesn't expire. Unused credit from a pack sits in your balance until you spend it on any supported model, not just Claude models.
Is this an official Anthropic reseller?
No. Token Master resells access to model capacity at a discount from list price. We are not affiliated with or endorsed by Anthropic, and nothing on this page should be read as an official partnership claim.
Sources
See /tm/about/ for how Token Master's resale model works and /tm/payments/ for pack and billing details.