langchain-opedd

Licensed, rights-cleared content for your LangChain pipelines. Opedd is the licensing rail between expert publishers and AI products — every article comes with a verifiable license key, on-chain proof (Tempo), and EU AI Act Article 53 attestation support. This package is the licensed alternative to scraping for RAG, agents, and AI search.

pip install langchain-opedd

Load your licensed corpus into RAG (the 3-line on-ramp)

from langchain_opedd import OpeddFeedLoader

loader = OpeddFeedLoader(access_key="ent_...")   # your Opedd enterprise access key
docs = loader.load()                              # LangChain Documents, licensing provenance in metadata

Every Document.metadata carries id, title, source, publisher_id, published_at, author, content_hash, and licensed: True — so answers in your pipeline stay attributable to licensed sources. Supports since (delta feeds), max_documents, and lazy_load() for streaming.

# Straight into a vectorstore:
from langchain_openai import OpenAIEmbeddings
from langchain_community.vectorstores import FAISS

vectorstore = FAISS.from_documents(docs, OpenAIEmbeddings())

Agent tools

from langchain_opedd import (
    OpeddLookupTool,          # licensing status + price for any article URL (no auth)
    OpeddDirectoryTool,       # browse verified publishers (no auth)
    OpeddVerifyLicenseTool,   # verify a license key + on-chain proof (no auth)
    OpeddContentTool,         # retrieve licensed article text (buyer token)
)

tools = [
    OpeddLookupTool(),
    OpeddDirectoryTool(),
    OpeddVerifyLicenseTool(),
    OpeddContentTool(buyer_token="opedd_buyer_live_..."),
]
# use with create_react_agent / your agent framework of choice

Where keys come from

• No key needed for discovery/verification tools.

• Buyer token (opedd_buyer_live_*): self-serve signup at opedd.com — no approval step.

• Enterprise access key (ent_*): issued with an order. What the loader does depends on the licence:

  • AI training and Full catalogue orders: the feed carries the full text; nothing else is needed.
  • Monthly AI answers and client display: the feed lists articles without their text, and the loader fetches each article's text from the content API. Pass buyer_email (the email on the order) so the loader can exchange your access key for a bearer token, or pass buyer_token if you already have one.
  • Pay per request (metered): the same, but every fetch is billed and returns a snippet (up to 300 words or 25% of the article).

  Pass allow_discovery_only=True to load metadata-only Documents and fetch nothing. Articles the content API will not serve (revoked, withdrawn by the publisher) are skipped.

loader = OpeddFeedLoader(access_key="ent_...", buyer_email="you@company.com")

Autonomous purchasing

Deliberately not a LangChain tool (payment confirmation belongs in a richer protocol). Agents that buy licenses mid-task should use the Opedd MCP server — hosted at https://mcp.opedd.com/mcp or local via npx opedd-mcp — which includes purchase_license with Stripe support.

Related

• Opedd MCP server — 17 tools for MCP-native agents (Claude, Cursor, OpenAI)
• opedd (Python SDK) — the client this package wraps
• For AI agents — full API documentation

MIT.