KeeperHub Agent SDK
Multi-framework KeeperHub SDK — let AI agents transfer, swap, and automate onchain.
Project Description
KeeperHub AI SDK is a multi-framework plugin system that enables AI agents to execute real on-chain actions using plain English.
We built a unified SDK that works across OpenClaw, LangChain (TypeScript & Python), and ElizaOS, allowing agents to interact with DeFi protocols, wallets, workflows, and on-chain data seamlessly.
The SDK exposes 30+ tools covering transfers, swaps, ENS, workflows, and automation, with access to 396 DeFi actions across 19+ chains including Base, Ethereum, Arbitrum, and Polygon.
Developers can get started quickly:
OpenClaw: npm install -g openclaw openclaw plugin install @ethglobal-openagent/openclaw-keeperhub openclaw
TypeScript (LangChain): npm install @ethglobal-openagent/langchain-keeperhub
Python (LangChain): pip install keeperhub-langchain
ElizaOS: npm install @ethglobal-openagent/elizaos-keeperhub
With just a few lines of code, developers can build AI agents that read data, generate workflows, and execute real on-chain transactions.
KeeperHub AI SDK makes it simple to connect AI agents to blockchain execution — turning natural language into real DeFi actions.
How it's Made
We built KeeperHub AI SDK as a unified TypeScript SDK on top of the KeeperHub REST API, which handles on-chain execution, workflows, and DeFi actions. The SDK abstracts API calls, adds retry logic, safety controls (like testnet-only mode), and exposes them as structured tools.
On top of this core layer, we created integrations across multiple agent frameworks:
LangChain (TypeScript & Python) using toolkits that expose 30+ structured tools OpenClaw plugins by wrapping LangChain tools as native CLI tools ElizaOS plugin with actions and providers mapped to KeeperHub capabilities
We also built a Telegram bot interface using the LangChain toolkit + OpenRouter LLM to demonstrate real-world usage.
A key design decision was creating a single shared tool layer, allowing the same capabilities to work across all frameworks without rewriting integrations. This makes the system modular, extensible, and easy to adopt.
