# Curtly

> Deterministic prompt and context compression for production LLM pipelines. Save 40-60% tokens with zero syntax drift.

- [Introduction](https://docs.curtly.dev/index.md): Welcome to your project
- [Quickstart](https://docs.curtly.dev/quickstart.md): Begin with a guide on the fastest path to a successful outcome
- [Compression Modes](https://docs.curtly.dev/architecture/compression-modes.md): Compare Conservative, Balanced, and Aggressive operational modes.
- [Immutable Safe Vault](https://docs.curtly.dev/architecture/safe-vault.md): Deterministic preservation mechanics for code blocks, JSON schemas, and templating tags.
- [POST /api/v1/compress](https://docs.curtly.dev/api-reference/compress.md): HTTP endpoint for prompt and context compression.
- [Client Integrations](https://docs.curtly.dev/integrations/sdks.md): Client wrappers and middleware examples for OpenAI SDK and Vercel AI SDK.
- [Security and Access Control](https://docs.curtly.dev/security/access-control.md): Data retention policy, transport encryption, and API key authorization scopes.