Settings reference
Every Contexta setting explained: API key source, Claude model, SEO plugin, token limits, rate limiting, instant indexing, Google submission and audit scope.
Settings reference
Contexta → Settings has four tabs: General, Import GSC, Memory and Activation. This page covers General and the advanced options.
API Key
Your Anthropic API key, from console.anthropic.com.
Once saved it is masked — paste a new key only to replace it. The remove button clears it and disables AI features until you add another.
API key source (Advanced)
On WordPress 7.0+, an Anthropic key can live in Settings → Connectors and be shared by
every plugin on the site. Contexta detects a registered anthropic connector automatically.
| Source | Behaviour |
|---|---|
| Auto (default) | Use the Connectors key when available, otherwise the plugin key |
| Connectors only | Only ever use the site-wide Connectors key |
| Plugin key only | Ignore Connectors and use the key stored here |
The panel shows which source is currently active, so there's no ambiguity about which key is being billed.
Claude model
Which model runs your generations. Current generation first, with the published API price per million tokens — which is what the editor's cost readout uses:
| Model | Best for | Input / output per MTok |
|---|---|---|
| Claude Fable 5 | Most capable, widely released | $10 / $50 |
| Claude Opus 5 | Complex reasoning and long tasks | $5 / $25 |
| Claude Sonnet 5 | Best balance of speed and intelligence — the default | $2 / $10 |
| Claude Haiku 4.5 | Fastest and cheapest — good for large batch runs | $1 / $5 |
| Claude Mythos 5 | Project Glasswing, limited access | $10 / $50 |
Previous-generation Opus 4.8, 4.7 and 4.6 and Sonnet 4.6 are also listed, along with a dated Haiku snapshot, for reproducibility.
If you saved a model that has since left this list, Contexta keeps sending it — your setting is not overwritten. Newly released model IDs are accepted too, before the dropdown catches up.
Pick a model your API key actually has access to. If you see "model not found", try Sonnet 5 or Haiku 4.5 — Fable 5, Opus 5 and Mythos 5 may require higher-tier access on your Anthropic account.
Thinking effort
How long Claude thinks before it answers. Thinking is billed as output tokens, and it shares the token limit with the answer itself.
That second part is the one that bites. If a model spends 500 tokens thinking about a 600-token SEO title, the answer arrives truncated — which is where a lot of "AI returned incomplete content" comes from on frontier models.
| Mode | What it does |
|---|---|
| Balanced (default) | Short tasks — titles, descriptions, scores — run at low effort. Content writing keeps full depth |
| Economy | Low effort everywhere. Cheapest and fastest; sensible for large batch runs |
| Maximum | Full depth on everything. This is exactly how Contexta behaved before this setting existed |
Claude Haiku 4.5 does not support this control at all, and neither do older models. Contexta simply omits the setting for them rather than sending something that would fail, and the settings screen tells you when your selected model ignores it.
Balanced is the right default for almost everyone: a meta description does not get better for being pondered, and you pay for the pondering either way.
SEO plugin
Where Contexta writes titles and meta descriptions. Auto-detect is the default and handles almost everyone.
Supported: Yoast SEO, Rank Math, All in One SEO, SEOPress, The SEO Framework, Slim SEO, or None (WordPress only).
The panel shows which plugin was detected, and warns you if the one you selected manually isn't active.
AI rate limit
Default: 0, meaning no cap.
Caps how many AI actions each logged-in user can run in an hour. 0 switches the cap off
entirely, and that is the default — a single-author site does not need to be rate-limited
against itself.
Set a number on a multi-author site, where it stops one enthusiastic editor from running up a large bill in an afternoon. The counter is a fixed hour per user: it resets on the hour from a user's first action, not on their last, so somebody working steadily still gets their allowance back.
This is a convenience guard, not your real safety net. A spend limit set in the Anthropic console is the one nothing can exceed — set that regardless of what you put here.
Instant indexing
On by default.
When you save SEO fields or content, Contexta notifies Bing, Yandex, Naver and Seznam through the IndexNow protocol so changes get re-crawled sooner.
Completely automatic — no accounts, no keys, no extra plugins. The verification key is generated and served virtually, so nothing is written to disk and it works on locked-down hosts.
Google does not support key-less instant submission; its sitemap is refreshed by WordPress core automatically. See Indexing for the full picture, including how to read Google's own verdict on each page.
A page marked noindex is never submitted, whatever this setting says.
Audit scope
Two audits fetch your own pages over HTTP, so how many they cover is a time trade-off.
| Setting | Where | Default |
|---|---|---|
| How many pages | AI Visibility → Speed | 20 (choose 5, 10, 20, 50 or 100) |
| Content audit sweep | AI Visibility → Content | Automatic, adjustable via rm_audit_max_posts |
Both prioritise your busiest pages from your Search Console import, so a small sample still covers the pages that matter. Every limit is filterable — see Developers.
Token limits (Advanced)
| Setting | Default | Controls |
|---|---|---|
| SEO token limit | 600 | Max output for title and meta generation |
| Content token limit | 2500 | Max output for intros, sections, FAQs, rewrites |
Higher values allow longer output but cost more per generation. The defaults suit most sites; raise the content limit if full rewrites are being cut short.
Submit pages to Google (Indexing API)
Optional, and the only feature with real setup. It pushes updated pages straight to Google rather than waiting for a normal crawl.
One-time setup:
- In Google Cloud, pick a project and enable the Indexing API.
- Create a Service Account, add a JSON key, and download the file.
- In Search Console → Settings → Users and permissions, add the service-account email as an Owner.
- Paste the whole JSON file into the field in Contexta.
The JSON is validated on save — if client_email or private_key are missing, Contexta
rejects it and leaves your existing credentials untouched.
Daily submission limit — default 200. Google's default quota is 200 URLs per day per project. Lower it if you share the project with other tools. The screen shows how many you have used today.
Once configured, enable Allow submitting pages to Google from the editor to turn on the button in the editor's Submit to Google tab.
Import GSC
Covered in Getting started.
Import the Dates file alongside Pages and Queries. It carries the period your export covers, which is what makes two imports comparable — without it a 28-day export next to a three-month one looks like a traffic collapse. Its name follows your Search Console language, and the import screen labels each field with the file name your account actually writes.
Indexing (Pro)
Connecting your own Google account to read index status is documented on the Indexing page.
Memory
Covered in Site memory. The tab appears only in Contexta Pro.
Activation
Your licence key or Envato purchase code, activation status, site count and support period. See Getting started.
Uninstalling
Deleting the plugin through WordPress removes all Contexta data: settings, imported Search Console data, memory layers, caches, token logs, AI traffic history and post meta. Deactivating alone changes nothing — your data waits for you.
If both the free plugin and Pro are installed, deleting one of them removes only that folder. The shared data is left alone for the copy you keep, and is cleared only when the last one is deleted.
On a multisite network, deleting the plugin clears every site on the network, not just the main one.
Multisite
Contexta stores its settings per site. Each site on a network has its own API key, Search Console imports, fix history and memory — which is what you want, since each site is its own property in Search Console with its own data.
That does mean configuring each site separately. Site administrators can do this themselves:
the screens need manage_options, and the editor needs edit_posts.