KeeperKit

KeeperKit is a comprehensive TypeScript SDK for the KeeperHub blockchain automation protocol, designed to give developers a fully-typed, ergonomic interface for managing workflows, triggering executions, and querying chains without the friction of hand-rolling raw HTTP calls.

To prove the power of the SDK, we built a full stack ecosystem around it, including an ElizaOS plugin for seamless agentic integration, comprehensive web documentation, and a consumer-centric dApp called "KeeperUniversity."

KeeperUniversity bridges the gap between decentralized finance and institutional academia, solving a real-world pain point we face as students: fragmented crypto liquidity. Students often have funds scattered across different chains and protocols (e.g., ETH on Aave, USDC in a Phantom wallet, staked SOL on Marinade). Our AI-powered assistant, KeeperAgent, allows a user to describe their intent in plain English. The agent then automatically builds a workflow that harvests funds from every source, swaps and bridges them as needed, and deposits the exact required fee into the university's endowment wallet.

For the administration side, the platform automates the routing of collected funds into designated channels (e.g., 30% to hostel management, 25% to lecture halls, 20% to mess/dining services) using transparent, on-chain execution with full audit trails. From base SDK to consumer UI, KeeperKit makes intent-driven blockchain automation accessible by making KeeperHub a self sufficient protocol in supplying everything to devs to build wonderful consumer apps.

We built this project across three distinct layers: the core infrastructure, the agentic integration, and the consumer application.

1. The Infrastructure (KeeperKit SDK): We engineered the SDK in TypeScript to act as the stable boundary between a host application and KeeperHub. It abstracts complex operations into 9 core modules with built-in retry mechanics, polling, and fallback handling. It handles resolving intent, loading workflow graphs, and executing operations (via client.executeWorkflow(id)). We also generated comprehensive developer documentation that outlines installation, distribution builds, and client initialization.

2. The Agentic Layer: To make the SDK actionable via AI, we built a custom plugin for ElizaOS. This allows autonomous agents to natively authenticate and interact with KeeperHub. We wrapped this into "KeeperAgent," an on-chain assistant that translates plain English prompts into executable directed workflow graphs composed of trigger nodes, condition nodes, and action nodes.

3. The Application Layer (KeeperUniversity): We built a modern web frontend to act as the consumer interface. The most complex technical hurdle was engineering the "Student Fee Collection" workflow. When a user submits an intent to pay, the system has to dynamically aggregate multi-chain positions. The KeeperKit-backed agent maps out the logic required to harvest yield (like Aave or Marinade), validate local topological flow, route the swaps, and handle the bridging. Once the workflow is validated safely by the agent, it executes the payload on KeeperHub, polls the status, and returns a shareable, readable execution ID to the university's finance office. By utilizing our own SDK to build the app, we dogfooded our own developer tooling to ensure the API surface is highly ergonomic.