langchain-novada 0.1.0

LangChain tools for Novada — web search, scrape, crawl, extract, research & proxy — derived LIVE from the Novada MCP server (zero contract drift).

langchain-novada

LangChain tools for Novada — web search, scrape, crawl, extract, research & proxy for your agents.

Tools and their input schemas are derived live from the Novada MCP server's tools/list, so this package never hard-codes the Novada API contract. When the server adds or changes a tool, your tools reflect it on the next call — zero drift.

Install

pip install langchain-novada

Quickstart (hosted — no install of the server)

from langchain_novada import aget_tools

# api_key from the argument or the NOVADA_API_KEY env var
tools = await aget_tools(api_key="nvd_...")

# hand them to any LangChain / LangGraph agent
from langgraph.prebuilt import create_react_agent
agent = create_react_agent("anthropic:claude-sonnet-4-6", tools)

Synchronous code:

from langchain_novada import get_tools
tools = get_tools(api_key="nvd_...")

Local mode (self-host the server via npx)

tools = await aget_tools(api_key="nvd_...", mode="local")  # spawns `npx -y novada-mcp`

Tool groups

The default hosted endpoint exposes the core group: novada_search, novada_extract, novada_crawl, novada_map, novada_research, novada_scrape, novada_verify. Request additional groups (e.g. proxy, platform scrapers) with groups=:

tools = await aget_tools(api_key="nvd_...", groups=["proxy", "scrape"])

Auth

Pass api_key= or set NOVADA_API_KEY. Get a key at https://dashboard.novada.com/overview/scraper/api-playground/.

Why this design

A LangChain wrapper that hand-copies an API contract drifts and breaks. This package instead asks the live Novada MCP server what tools exist and what they accept, so the contract has a single source of truth (the server) and a CI live contract test that fails the build if anything drifts.

License

MIT © Novada