init() patches the LLM provider SDKs you already use so every call is
scanned before it is sent (auto-instrumentation),
and a Guard client calls the Guard API directly when you
want to scan text yourself. Both are open source on GitHub.
Get started
Quickstart
Get an API key, install the SDK and see a request blocked, in about five
minutes.
How-to guides
Python SDK
Install, initialise, the Guard client, error handling and retries.
Node.js SDK
The same for Node.js, plus the Vercel AI SDK middleware.
Frameworks
LangChain, LangGraph, LlamaIndex, CrewAI, FastAPI and the Vercel AI SDK.
Migrate from OpenAI
Move an OpenAI integration onto the proxy instead of the SDK.
Reference
Python SDK reference
Every public class, function and method, generated from the published package.
Node.js SDK reference
The same for the npm package.
Auto-instrumentation
Which provider calls
init() patches, which it does not, and how to check.Error handling
Every error code the API returns, and how to handle each one.
Upgrading
Upgrade to SDK and CLI 2.0
What 2.0.0 removed, what to use instead, and how to find the calls.
Versioning and support
How long each SDK major version gets fixes.
Troubleshooting
Troubleshooting
Authentication, connection, blocked-request and streaming problems, with the
fix for each.