Question 1

What is AgentRun?

Accepted Answer

AgentRun is an open-source workflow DSL designed to turn repeatable parts of AI agent work into structured, executable workflows. It allows combining tool calls, code, decisions, and agent invocations in a single workflow definition.

Question 2

How does AgentRun relate to agents?

Accepted Answer

AgentRun works with the agents you already run. Instead of replacing your agent, it adds orchestration on top, letting you define workflows that call tools, make decisions, and invoke your agent only when needed.

Question 3

What is Jev in AgentRun?

Accepted Answer

Jev powers the decision nodes in AgentRun, enabling fast, focused 'system one' decisions for tasks like routing, screening evidence, and validating criteria. The workflow code reads the decision and its confidence to choose the next step.

Question 4

Can I try AgentRun without an API key?

Accepted Answer

Yes. The repository includes scripted demo examples that run without any API key. For example, `npm run demo:support` runs the support workflow with scripted tools, Jev answers, and agent responses, printing a report for each case.

Question 5

How do I write a workflow in AgentRun?

Accepted Answer

You can write workflows as JSON or use the TypeScript builder with Zod contracts. The builder infers input and output types, and the interpreter checks intermediate state paths at runtime.

AgentRun

Alternatives

GitHub Copilot

Cursor

Replit AI

v0 by Vercel

Bolt.new

Codeium