agent-express
Lightweight TypeScript framework for AI agents using familiar Express-style middleware.
View on GitHub Updated 2026-06-15
What is agent-express?
Agent-express is a minimalist middleware framework designed specifically for building AI agents in TypeScript. It applies the well-known Express.js (ctx, next) pattern to agent workflows, reducing the learning curve for backend developers who already understand middleware stacks.
The system organizes behavior through five onion-style hooks that let middleware run before and after key stages such as session start, model calls, and tool execution. Built-in options cover cost limits, validation, logging, token tracking, and context compaction without requiring separate libraries.
It targets TypeScript developers who want a simple, composable way to create reliable agents rather than learning dozens of new abstractions. The framework works with any provider supported by the AI SDK and ships with CLI tools for local development and testing.
What you can build with agent-express
Production support agent
Combine budget guards, PII redaction, and RAG search middleware to create a safe customer support bot that escalates when needed.
Multi-model routing setup
Use the router middleware to automatically select cheaper or more capable models based on query complexity while tracking token usage.
Custom analytics layer
Write a small middleware object that records turn counts and model costs across sessions for internal reporting dashboards.
Install agent-express
Install
npm install agent-expressQuick start
npm install agent-express- 1Run npm install agent-express to add the package.
- 2Import Agent and any middleware namespaces you need.
- 3Create a new Agent instance with your chosen model and instructions.
- 4Attach middleware such as tools or guards using the use() method.
- 5Call agent.run() with your input to execute the agent.
agent-express: pros & cons
Pros
- +Very small conceptual surface area thanks to the single middleware pattern.
- +Strong TypeScript support and Zod-based structured output validation.
- +Flexible observability and testing tools included out of the box.
- +Easy to extend with custom middleware for almost any behavior.
Cons
- –Requires TypeScript and familiarity with middleware concepts.
- –Fewer turnkey agent templates compared with larger frameworks.
- –Documentation and community are still growing.
Did you find this helpful?
Frequently asked questions
Yes, any provider supported by the AI SDK can be used by passing a provider/model string when creating the agent.
User reviews
Verified reviews from the community shape this listing's rating.
Loading reviews…
Similar agents
Other general-purpose options worth comparing.
