Skip to main content
PlainID provides policy-based authorization (PBAC) and centralized policy enforcement. This integration allows you to enforce fine-grained access control within your LangChain applications at different stages of the LLM chain. With langchain-plainid, you can:
  • Filter RAG data: Dynamically filter documents retrieved from your vector store based on the user’s permissions, ensuring they only see data they are authorized to access.
  • Authorize prompts: Control whether a user or tenant is allowed to invoke a chain or tool based on the category of their query.
  • Anonymize data: Detect and anonymize (mask or encrypt) PII or other sensitive entities in responses, based on policies defined in PlainID.

​
Installation

First, install the partner package: 
pip install langchain_plainid

​
Setup

Next, you need to configure the provider with credentials from your PlainID tenant. You will need your Client ID, Client Secret, and Base URL. You can set these as environment variables: 
export PLAINID_CLIENT_ID="your-client-id"
export PLAINID_CLIENT_SECRET="your-client-secret"
export PLAINID_BASE_URL="[https://platform-product.us1.plainid.io](https://platform-product.us1.plainid.io)"
All components will automatically use these environment variables if no explicit credentials are provided during instantiation.

​
Usage

The package provides three main components for enforcing authorization.

​
RAG Data Filtering

The PlainIDRetriever wraps your existing vector store retriever. It fetches authorization filters from PlainID based on the user’s identity and applies them to the vector store query. This filters out documents before they are passed to the LLM for context. This example assumes you have a PlainIDPermissionsProvider configured (e.g., via environment variables) and a PlainIDRetrieverFilterProvider set up.

​
Prompt Authorization

The PlainIDCategorizer can be placed at the beginning of a chain to authorize the user’s intent. It classifies the input prompt (e.g., “HR”, “Finance”, “Contract”) and checks with PlainID if the user is permitted to ask about that category. If not authorized, it raises a ValueError.

​
PII Anonymization

The PlainIDAnonymizer can be placed at the end of a chain to inspect the LLM’s response. It uses presidio to detect PII entities (like “PERSON”, “PHONE_NUMBER”) and then consults PlainID on whether to MASK or ENCRYPT them based on defined policies. For more detailed information and full examples, refer to the langchain_plainid PyPI page.
Edit this page on GitHub or file an issue.
Connect these docs to Claude, VSCode, and more via MCP for real-time answers.