K8 Inspector uses your own Anthropic API key for AI features. Understand exactly what it costs and how to control spending.
| Model | Input Tokens | Output Tokens | Used For |
|---|---|---|---|
| Claude Sonnet 4 | $3.00 / 1M tokens | $15.00 / 1M tokens | Complex analysis, RCA, debugging |
| Claude Haiku 3.5 | $0.80 / 1M tokens | $4.00 / 1M tokens | Simple queries, listing, filtering |
Source: anthropic.com/pricing - Always verify current rates before relying on these estimates.
| Operation | Example | Est. Tokens | Est. Cost |
|---|---|---|---|
| Simple query | "List failing pods in production" | ~700 total | ~$0.006 |
| Complex query | "Why is checkout-service slow?" | ~3,000 total | ~$0.02 |
| Root cause analysis | Automated incident diagnosis | ~7,000 total | ~$0.05 |
| Full cluster diagnosis | "What's wrong with my cluster?" | ~10,000 total | ~$0.08 |
| YAML generation | "Create a deployment for nginx" | ~2,300 total | ~$0.02 |
| Security explanation | CVE context and remediation | ~4,500 total | ~$0.03 |
Example: A complex query with 2,000 input tokens and 1,000 output tokens:
(2,000 x $0.000003) + (1,000 x $0.000015) = $0.006 + $0.015 = $0.021
K8 Inspector provides built-in controls to help manage your API spending.
| Control | Default | Description |
|---|---|---|
| Queries per minute | 10 | Prevents runaway automated queries |
| Queries per hour | 100 | Hourly usage cap |
| Queries per day | 500 | Daily maximum queries |
| Max tokens per query | 8,000 | Limits context sent to AI |
| Monthly budget cap | Unlimited | Disables AI when budget reached |
"Show failing pods in namespace payments" costs less than "Show me everything wrong." Targeted queries require less context.
Configure K8 Inspector to use Claude Haiku for list/filter operations. It's 75% cheaper than Sonnet for basic queries.
Repeated similar queries can use cached context, reducing token usage and API calls.
Configure a monthly budget cap to automatically disable AI features before costs exceed your threshold.
For non-critical dev clusters, disable automatic AI analysis and use manual triggers only.
Track your actual usage in the Anthropic Console. Set up billing alerts directly with Anthropic.
Yes. All core Kubernetes management features work without an API key. AI-powered features (assistant, auto-analysis, intelligent insights) will be disabled, but dashboards, logs, resource management, and security scanning work fully.
Sign up at console.anthropic.com to create an Anthropic account and generate an API key. You'll need to add payment information to your Anthropic account.
When you use AI features, cluster metadata (resource names, statuses, events, log snippets) is sent to Anthropic for analysis. Secret contents are never sent. You can further restrict what's shared via configuration. Review Anthropic's data policies for their handling practices.
Anthropic applies rate limits based on your account tier. If you hit their limits, queries will fail temporarily. Consider upgrading your Anthropic account tier for higher limits, or configure K8 Inspector's rate limits to stay within your allocation.
Currently, K8 Inspector is optimized for Anthropic Claude. Support for additional providers may be added in future releases. Check our documentation for the latest supported providers.
Monitor your usage in K8 Inspector's Settings > AI Usage panel for estimates. For authoritative billing information, check the Anthropic Console directly. We recommend setting up Anthropic billing alerts.
Download K8 Inspector free. Add your Anthropic API key when you're ready to unlock AI features.