Choosing a Framework

Compare OpenClaw and AgentKit to find the right foundation for your AI agent on Base

Your first decision when building an AI agent on Base is choosing a framework. This determines how you write your agent’s logic, where it runs, and how much infrastructure you manage yourself.

Comparing your options

	OpenClaw	AgentKit
What it is	An open-source AI assistant that runs locally or on a server, extended with skill plugins — skills also work in Claude Code, Codex, and OpenCode	A developer toolkit for building custom agents with full control over behavior and integrations
Language	TypeScript (plugins); any LLM via config	TypeScript / Python
Hosting	Self-hosted (local or VPS)	Self-hosted
Skill system	Plugin-based skills installed via CLI — compatible with Claude Code, Codex, and OpenCode	Tool definitions you write in code
Best for	Multi-channel assistants with pre-built skills and a CLI-first workflow; also great for adding onchain skills to an AI coding tool	Production agents where you need full control over logic, integrations, and deployment

OpenClaw

OpenClaw is an open-source AI assistant you install and run locally. It connects to multiple messaging platforms out of the box (Discord, Telegram, web) and comes with a plugin system for adding skills — pre-built actions your agent can perform onchain. You configure it through a config.json file and extend it by installing skill packages.

Use OpenClaw when you want to run a conversational agent across multiple channels and extend it with community-built or custom plugins.

Claude Code, Codex, and OpenCode are compatible — All skills and wallet tools built for OpenClaw work with AI coding tools too. Install any skill package with npx skills add inside Claude Code, Codex, or OpenCode and the onchain capabilities are available immediately. You can use whichever tool fits your workflow.

Get started with OpenClaw →

AgentKit

AgentKit is Coinbase’s developer toolkit for building agents that can interact with onchain services. You write the agent logic, define its tools, and deploy it on your own infrastructure. This is the most flexible option: you control every aspect of the agent’s behavior, from how it processes prompts to which services it calls.

Use AgentKit when you’re building a production agent and want full control over its architecture in TypeScript or Python.

Get started with AgentKit →

How to choose: Use OpenClaw if you want a multi-channel chatbot with a plugin-based skill system and minimal boilerplate. Use AgentKit if you need full control over your agent’s logic, tool definitions, and deployment pipeline.

Next steps

OpenClaw quickstart

Install OpenClaw and send your first onchain transaction in minutes.

AgentKit quickstart

Set up AgentKit and build your first agent with a CDP wallet.