PII never reaches your LLM. One line of code.

Armos wraps the OpenAI and Anthropic SDKs to automatically detect and mask personally identifiable information before it leaves your server — and restore real values in the response. Your application code changes by exactly one word.

# Before
client = OpenAI()

# After
client = ArmosOpenAI(OpenAI())

Detection runs entirely on your machine using Presidio + spaCy. No data sent to any Armos server — there is no Armos server.

pip install armos
python -m spacy download en_core_web_lg

→ armos-python — full docs, quickstart, and benchmarks
→ armos.dev — landing page