# AITHURA

> AITHURA is one control layer for every AI action: a zero-trust, BYOK (bring your own key) security gateway that sits between applications and LLM providers, AI agents, MCP servers, and tools. It secures AI activity, governs agent actions, and produces audit-grade evidence of what happened — hosted, private VPC, self-hosted, or air-gapped.

AITHURA is not an LLM provider. Customers keep their own OpenAI, Anthropic, Google, or OpenAI-compatible keys (stored encrypted with AES-GCM-256). Every request and every agent tool call is inspected inline by the Sentinel Protocol — 81 deterministic security and governance engines — before it reaches the upstream model or tool. Dangerous prompts (e.g. high-confidence prompt injection) are blocked before they reach the provider.

## Key facts

- Category: AI security gateway / LLM firewall / agent and MCP security / AI governance and compliance evidence.
- Outcomes: Secure (block prompt injection, data exfiltration, PII leakage, unsafe output), Govern (per-tool allow / warn / block policies for agents), Prove (signed action receipts, AIBOM, forensic replay, exportable audit logs).
- Engine: Sentinel Protocol, 81 deterministic engines, open source, self-hostable as a sidecar on port 8787.
- Integration: OpenAI-compatible — change the base URL and add an `x-aithura-key` header (or `Authorization: Bearer <AITHURA key>`). AITHURA keys start with `aithura_live_`.
- Agent Shield: a gateway for AI agent tool calls; unknown tools are auto-discovered in `warn` mode; explicit tool modes override automated risk scores.
- Frameworks mapped: OWASP LLM Top 10, NIST AI RMF, SOC 2 evidence, EU AI Act traceability, NDAA 1513, NIST SP 800-53, NIST SP 800-171, DFARS 252.204-7012, CMMC Level 2, DoD IL4/IL5 path.
- Data minimization: prompts and responses are truncated to 100 characters in stored logs.

## Pages

- [Home](https://aithura.com/): Platform overview, outcomes, deployment options, and assurance mapping.
- [Docs](https://aithura.com/docs): Public API documentation — LLM gateway, Agent Shield, keys, and quick start.
- [Agents](https://aithura.com/agents): Agent Shield and the Sovereign Agent Guardrail for autonomous agents and MCP.
- [Defense & Intelligence](https://aithura.com/defense): Air-gapped deployment, ATO readiness, NDAA 1513, CMMC, chain of custody.
- [Secure OpenClaw](https://aithura.com/secure-openclaw): Run OpenClaw behind the local Sentinel sidecar.
- [Support](https://aithura.com/support): Contact and help.

## Optional

- [Full reference for LLMs](https://aithura.com/llms-full.txt): Detailed integration, engines, and compliance reference.
- [GitHub](https://github.com/AITHURA-com/aithura): Source and Sentinel Protocol.