Skip to main content

Getting Started

Start building governed, context-aware LLM applications in minutes. The CRP SDK gives you unbounded context, automatic continuation, grounded retrieval, and built-in safety governance - all through a single crp.SDKClient() interface.

Deployment status

Managed cloud CRP Gateway and CRP Comply are on the waitlist at gateway.crprotocol.io and comply.crprotocol.io. Self-host the SDK today via pip install crprotocol.

• Installation

  Install CRP with pip and choose the extras that match your stack.

• CLI Reference

  Run python -m crp commands for shell-based workflows.

• Providers

  Configure OpenAI, Anthropic, Ollama, or custom providers.

• Local Models

  Run CRP with LM Studio, Ollama, llama.cpp, or vLLM - no cloud required.

The fastest path to value

import crp

client = crp.SDKClient()              # auto-detects provider from env
response = client.complete("Explain CRP in one paragraph.")

print(response.text)
print(f"Risk: {response.crp.risk}")
print(f"Grounded: {response.crp.grounded}")

That's it - one client, one call, and CRP handles headers, safety, provenance, and context packing behind the scenes.