AgentDock
AgentDock is a TypeScript framework for building AI agents with configurable determinism, balancing creative LLM behavior with predictable workflows. It includes AgentDock Core, a backend-first, framework-agnostic core, and an open-source Next.js client. It is for developers who need reliable, production-ready AI agents that can operate with minimal supervision.
✨ Key features
- Node-based architecture with BaseNode and AgentNode
- Configurable determinism: mix LLM creativity with deterministic workflows
- Framework-agnostic and provider-independent design
- TypeScript with comprehensive type safety
- Includes example agents like Dr. House and Cognitive Reasoner
- Advanced capabilities: sessions, orchestration, storage, evaluations
🎯 Use cases
- Build diagnostic agents that orchestrate search and research tools
- Create multi-stage reasoning engines for complex problem-solving
- Develop educational agents with vectorized knowledge and diagram rendering
- Implement vision-based nutritional analysis from food images
- Deploy custom AI agents with controlled predictability
📦 Installation
🧰 Requirements: Requires Node.js ≥ 20.11.0 (LTS), pnpm ≥ 9.15.0, and API keys for LLM providers (e.g., Anthropic, OpenAI).
git clone https://github.com/AgentDock/AgentDock.git
cd AgentDock
corepack enable
corepack prepare pnpm@latest --activate
pnpm install
For a clean reinstallation:
pnpm run clean-install
Create environment file:
cp .env.example .env.localcp .env.example .envAdd your API keys to the environment file.
Start development server:
pnpm dev
🚀 Usage
pnpm dev
After starting the development server, access the open-source client at the local URL (typically http://localhost:3000) and interact with the agents.
⚠️ Good to know
The storage system is evolving; vector storage and additional backends are in development.
❓ FAQ
What is configurable determinism?
It's a design principle that lets you control how predictable your agent is. You can use LLM inference for creative tasks and deterministic workflows for reliable steps, combining both as needed.
Can I use AgentDock Core independently?
Yes, AgentDock Core is a separate framework that can be used in your own applications, independent of the open-source client.
What are the system requirements?
You need Node.js 20.11.0 or higher, pnpm 9.15.0 or higher, and API keys for LLM providers like Anthropic or OpenAI.
Are there example agents included?
Yes, the repository includes several example agents like Dr. House, Cognitive Reasoner, History Mentor, and Calorie Vision.
📊 Repository
🤖 Overview, features, install steps and FAQ were generated from the project's README on Sep 4, 2026. Always check the original source before running commands.