Workspace:
Savings over time
Request volume
Cache opportunity
Cache status mix
Top providers / models
Savings by feature
Tag requests with metadata.feature (e.g. "content-moderation") and this table shows which parts of your product drive usage and savings. Untagged traffic groups under "(untagged)".
| Feature | Requests | Est. savings | Verified savings | Cache hits |
|---|
| Created | Provider | Model | Mode | Cache | Raw→Opt | Est. $ | Verified $ | Latency | Templates |
|---|
API Keys
Send as Authorization: Bearer <key>. The full key is shown only once, at creation.
| Name | Preview | Status | Created | Last used |
|---|
Provider Credentials
Connect the AI key that already powers your app — your provider keeps billing you directly, and KillToken uses the key only for the requests you send. Supported: OpenAI, Anthropic, Gemini, Mistral, DeepSeek, OpenRouter, Together, Perplexity, xAI, Azure OpenAI, any self-hosted OpenAI-compatible endpoint, AWS Bedrock, or Google Vertex AI. Keys/secrets are encrypted at rest and shown only once, at creation. KillToken never uses platform-owned provider keys, AWS env credentials, or Google ADC. Azure OpenAI, openai_compatible, AWS Bedrock, and Google Vertex also need their non-secret config (region/project).
| Provider | Label | Status | Preview | Config | Created | Last used |
|---|
Support & resources
Auto-refreshes every 20s. No raw prompt content is ever shown.