langchain-keeperhub

Give your LangChain agent the ability to send tokens, call smart contracts, and resolve ENS names — without ever touching a private key.

langchain-keeperhub is a Python SDK that connects LangChain / LangGraph agents to KeeperHub, a Web3 execution service. You provide an API key; KeeperHub handles wallet management and transaction signing inside a secure enclave (Turnkey TEE). Your code never sees the private key.

Originally built for ETHGlobal OpenAgents.

Independent community project by @devendra116. Not affiliated with or endorsed by KeeperHub.

What can it do?

Your agent gets these as LangChain tools. It decides when and how to use them based on the user's request.

Install

pip install langchain-keeperhub

Optional extras:

pip install "langchain-keeperhub[workflows]"  # MCP workflow tools
pip install "langchain-keeperhub[ens]"         # keccak fallback (if OpenSSL lacks it)

Quick start

1. Get a KeeperHub API key at app.keeperhub.com (starts with kh_).

2. Install dependencies:

pip install langchain-keeperhub langchain langchain-google-genai langgraph python-dotenv

3. Write your agent:

from dotenv import load_dotenv
from langchain.agents import create_agent
from langchain_google_genai import ChatGoogleGenerativeAI
from langchain_keeperhub import KeeperHubToolkit

load_dotenv()  # reads KEEPERHUB_API_KEY and GOOGLE_API_KEY from .env

toolkit = KeeperHubToolkit()  # or pass api_key="kh_..." directly
agent = create_agent(
    model=ChatGoogleGenerativeAI(model="gemini-2.5-flash", temperature=0),
    tools=toolkit.get_tools(),
)

result = agent.invoke({"messages": [("user", "Send 1 USDC on Base Sepolia to 0x...")]})
print(result["messages"][-1].content)

Any LangChain-compatible model works (OpenAI, Anthropic, etc.) — the examples just happen to use Gemini.

Tools reference

After any write tool (transfer_funds, contract_call, check_and_execute), always call get_execution_status with the returned execution_id to get the final transaction hash.

Safety options

toolkit = KeeperHubToolkit(
    api_key="kh_...",
    testnet_only=True,                       # block writes to mainnet chains
    allowed_chain_ids={"11155111", "84532"},  # only Sepolia + Base Sepolia
    history=True,                             # log every write to SQLite
)

Use without an LLM

You can also use KeeperHubClient directly for scripting — no agent needed:

import asyncio
from langchain_keeperhub import KeeperHubClient

async def main():
    async with KeeperHubClient() as client:
        chains = await client.list_chains()
        print(chains)

asyncio.run(main())

Environment variables

Compatibility

• Python 3.10+
• v0.4+ requires langchain-core 1.x. If you're on langchain-core 0.3.x, use langchain-keeperhub 0.3.x.

Development

git clone https://github.com/devendra116/langchain_keeperhub.git
cd langchain_keeperhub
pip install -e ".[dev]"
pytest

Docs

• Getting Started — install, configure, build your first agent
• Architecture — how the SDK works internally
• ENS Integration — ENS and Basenames resolution details

Links

• KeeperHub
• KeeperHub API docs
• Turnkey signer (blog post)
• Examples

License

MIT — see LICENSE.

Independent community package. Not an official KeeperHub release. Trademarks belong to their owners.