# KillToken

> KillToken is an adaptive LLM optimization gateway. You route your existing OpenAI/Anthropic/Gemini (and other) chat traffic through it using your OWN provider API keys (strict BYOK); KillToken measures cost, optimizes prompts, caches repeat requests, and reports savings per feature on a dashboard. Base URL: https://killtoken.io — all API routes live under /v1.

Key facts for integrators (humans and AI agents):

- Authentication: `Authorization: Bearer kt_...` using a KillToken tenant API key created in the dashboard. This is NOT a provider key. Server-side only — never ship it in browser or mobile code.
- Strict BYOK: your provider key (e.g. OpenAI `sk-...`) is stored encrypted in KillToken per tenant and used for the actual model call. Your provider still bills you directly. If no credential is stored for the requested provider, requests fail fast with `400 provider_credential_required`.
- Drop-in path for OpenAI SDK users: point the SDK at `baseURL: https://killtoken.io/v1/openai` with your `kt_` key — chat completions only, no streaming.
- Tag traffic per app feature with the `x-killtoken-feature` header (or `metadata.feature` in the body) to get a per-feature savings breakdown on the dashboard.
- Enable response caching per request with `cachePolicy: { "exactCache": "read_write" }` on repeat-safe calls.

## Docs

- [Full integration guide for AI agents and developers](https://killtoken.io/llms-full.txt): every endpoint, header, error code, and copy-paste example in plain markdown
- [Human-readable API docs](https://killtoken.io/docs)
- [Pricing and plans](https://killtoken.io/pricing)

## Optional

- [Security practices](https://killtoken.io/security)
- [Privacy policy](https://killtoken.io/privacy)
- [Terms of service](https://killtoken.io/terms)
- [Contact](https://killtoken.io/contact): support@killtoken.io