Costs and your API key

Costs and your API key

Contexta uses your Anthropic API key. You pay Anthropic directly for what you use. We never resell AI credits, never mark them up, and never proxy your requests through our servers.

That means two bills, not one: the plugin licence, and whatever Anthropic charges for the generations you run.

Getting a key

1. Create an account at console.anthropic.com
2. Add billing
3. Go to Settings → API keys and create a key
4. Paste it into Contexta → Settings → General

A key can be used across several sites. On WordPress 7.0+ you can store it once in Settings → Connectors and share it with every plugin — see Settings.

Seeing what you spend

Contexta tracks token usage for every action and shows it in three places in the editor's top bar:

• All — total tokens and cost across every article, since install
• Page — total for the article you have open
• Last action — what the generation you just ran cost

Costs are estimated from published Anthropic per-million-token rates. Treat them as close approximations for budgeting, and your Anthropic console as the source of truth for billing.

Roughly what things cost

Exact figures depend on the model you pick and how long your content is, but the relative ordering is stable:

Model choice is the biggest lever

The same generation costs twenty-five times more on the most capable model than on the cheapest. Published price per million tokens:

Claude Sonnet 5 is the default and the right choice for most SEO writing. Haiku 4.5 cuts cost further, which makes it sensible for large batch runs you will review anyway. Reach for Opus 5 or Fable 5 on genuinely hard work, not on routine meta descriptions.

Contexta prices each model separately, so the cost shown in the editor reflects the model you actually selected.

Keeping spend under control

Set a budget in the Anthropic console. This is the real safety net — a hard spend limit on your account that no plugin bug or enthusiastic colleague can exceed.

Set the rate limit on multi-author sites. It is off by default, since a single author does not need capping. Where several people can run AI actions, a per-user hourly cap is your protection against a surprise bill. See AI rate limit.

Use Balanced thinking effort. The default. Short tasks stop paying for deep thinking they do not need. See Thinking effort.

Lower the token limits. 600 for SEO and 2500 for content are the defaults. Reducing them directly reduces cost per generation.

Be deliberate with batch runs. Batch multiplies cost by the number of articles times the number of tasks. Run it on 5 articles first, look at the counter, then scale up.

Competitor research caches for 30 days. Don't re-run it on the same article unnecessarily — the findings are reused automatically by every later generation.

What is sent to Anthropic

Only what a generation needs, and only when you click something:

• The article's title, content and URL
• The page's ranking queries from your Search Console import
• Your site memory context block
• Competitor research findings, when they exist

There are no background API calls. Contexta does not generate anything on a schedule, on save, or on page load. Every request follows a click.

Your Search Console data stays in your WordPress database. Only the queries relevant to the page you are working on travel with a prompt.

If AI features stop working

• "No API key configured" — the key is missing or was removed
• "Model not found" — your account lacks access to the selected model; switch to Sonnet 5 or Haiku 4.5
• Rate limit message — you hit the per-hour cap; wait, or raise it in Settings
• Licence not active — AI endpoints are gated behind activation

More in Troubleshooting.