AML - Agent Markup Language
an asynchronous TypeScript and JSX runtime for composing provider-agnostic agent workflows.
FreeAIOpen SourceFramework285 impressionslaunches Jan 30
AML is an asynchronous TypeScript/JSX runtime for building provider-agnostic AI agent workflows.
- for
- Developers building composable, multi-step AI agents with TypeScript.
- pricing
- open source
- license
- MIT
- sandbox-v0.5.2sandbox v0.5.2a month ago
- v0.8.1sdk v0.8.12 months ago
- sandbox-v0.5.1sandbox v0.5.12 months ago
Key features
- JSX component model — Define agents, tools, sandboxes and workspaces as reusable JSX components.
- Provider-agnostic agents — Choose OpenCode, Codex, GitHub Copilot, GLM or Pi per agent node.
- Sandbox execution — Run agent code in local, Docker, Daytona or Modal sandboxes with automatic setup/cleanup.
- Durable workspaces — Persist files across runs using local folders or S3-compatible storage.
- Parallel branching — Execute independent branches concurrently with the <Parallel/> component.
- Structured tracing — Inspect turns, tool calls, timings and failures via built-in traces.
Use cases
- Run a code review agent that executes static analysis tools in parallel.
- Generate weekly stand-up summaries by chaining a data-gathering agent with a synthesis agent.
- Build a release-note generator that reads repository files from a persisted workspace.
AML - Agent Markup Language vs alternatives
 AML - Agent Markup Language |  Fine |  Pocketenv |  Fluent CI |  Daytona | |
|---|---|---|---|---|---|
| Best for | Composable JSX-based AI agent workflows | AI-driven software generation | General sandbox execution for agents | TypeScript CI/CD pipelines | Infrastructure-as-code sandbox setup |
| Pricing | Open source | Free | Free | Free | Free |
| DevHunt upvotes | 1 | 75 | 73 | 158 | 339 |
| Launched | Jan 2029 | Jan 2023 | Apr 2026 | Jan 2024 | May 2024 |
- AML - Agent Markup Language vs Fine: Focuses on building software with AI agents but lacks JSX component composition and sandbox/workspace management.
- AML - Agent Markup Language vs Pocketenv: Provides a universal sandbox runtime but does not include the JSX-based workflow orchestration that AML offers.
- AML - Agent Markup Language vs Fluent CI: Enables CI/CD pipelines in TypeScript, whereas AML orchestrates multi-step AI agent flows.
- AML - Agent Markup Language vs Daytona: Daytona offers sandbox provisioning; AML combines sandbox, agent, and workspace orchestration in one framework.
AML - Agent Markup Language FAQ
How do I install AML?+
Run `npm install @aml-jsx/sdk` in a Node.js 26+ TypeScript project.
Can I use my own AI model?+
Yes, you can define a custom agent provider via `defineAgentProvider()` and plug it into the JSX tree.
How are resources cleaned up?+
The AML runtime automatically acquires and releases sandboxes and workspaces after each evaluation.
Is concurrency supported?+
Yes, the <Parallel/> component runs independent branches concurrently and aggregates their results.
Where can I see execution details?+
AML provides structured traces showing agent turns, tool calls, timings and any failures.
Summarized by DevHunt from aml.wearesingular.com · Sep 30, 2026. Details may change; check the official site.
