# Gray Swan AI

> Gray Swan AI is an AI security company. Its Cygnal product is a real-time,
> drop-in secure proxy that fronts LLM providers (OpenAI, Anthropic, Gemini
> formats) with input/output filtering and threat monitoring for jailbreaks
> and prompt injection, governed by versioned enforcement Policies. Gray Swan
> also runs Shade (adversarial red-teaming), Arena (a global red-teaming
> network), and adversarial model evaluations used by frontier labs.

This llms.txt was generated by the API Evangelist enrichment pipeline from
the provider's public surface and the Cygnal OpenAPI. Generated 2026-07-19.

## API

- [Cygnal OpenAPI 3.1](https://api.grayswan.ai/openapi.json): Machine-readable spec. Base URL https://api.grayswan.ai
- [API Reference](https://docs.grayswan.ai/api-reference): Endpoint reference
- [Creating Completions](https://docs.grayswan.ai/cygnal/creating-completions): Quickstart for the proxy
- Auth: `grayswan-api-key` header + a `policy-id` (or `agent-id`) selector header — no default policy
- Rate limits (free tier): 200,000 tokens/min, 10,000,000 tokens/day; over quota proxies without filtering

## Endpoints (Cygnal)

- POST /cygnal/v1/chat/completions — OpenAI-compatible chat completions (filtered)
- POST /cygnal/responses — OpenAI Responses-compatible
- POST /cygnal/v1/messages — Anthropic Messages-compatible
- POST /cygnal/v1beta/models/{model}:generateContent — Gemini-compatible
- POST /cygnal/monitor — monitoring only, no inference proxying
- GET/POST /policies, GET/PUT/DELETE /policies/{policy_id} — policy CRUD
- GET /policies/{policy_id}/versions — policy version history
- POST /policies/import, /policies/bulk-import — policy import

## Products

- [Cygnal](https://www.grayswan.ai/): Runtime AI defense (the API above)
- [Shade](https://www.grayswan.ai/): Adversarial red-teaming platform
- [Arena](https://www.grayswan.ai/): Global red-teaming network

## Docs & Platform

- [Documentation](https://docs.grayswan.ai/)
- [Platform / Get API key](https://platform.grayswan.ai)
- [Blog](https://www.grayswan.ai/blog)
- [GitHub](https://github.com/GraySwanAI)

## Security & Trust

- [Security / Responsible Disclosure](https://grayswan.ai/security)
- [security.txt](https://grayswan.ai/.well-known/security.txt)
- [Trust Center](https://trust.grayswan.ai/): SOC 2 Type 2, Cyber Essentials

## Legal

- [Terms of Service](https://www.grayswan.ai/terms-of-service)
- [Privacy Policy](https://www.grayswan.ai/privacy-policy)
- [Acceptable Use Policy](https://www.grayswan.ai/acceptable-use-policy)