# openai/openai-agents-js

A lightweight, powerful framework for multi-agent workflows and voice agents

Repository: https://github.com/openai/openai-agents-js
Canonical: https://ross.abutalabs.com/products/openai-agents-js
Homepage: https://openai.github.io/openai-agents-js/
Language: TypeScript
License: MIT
License Family: permissive
Topics: agents, openai, openai-api, realtime-api, typescript
Last push: 2026-08-26T04:01:11+00:00

## Health v2 (maintenance only)
Score: 81/100 (v2, computed 2026-09-03T02:39:23.370411+00:00)
- activity 99, release rhythm 86, longevity 32
- inputs: {"age_days": 460, "days_push": 7, "days_rel": 14, "gap_med": 2, "n_releases_24m": 74}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 3714, forks 936 (observed 2026-08-28T04:08:15.183253+00:00)

## What it is
A lightweight, provider-agnostic TypeScript/JavaScript framework from OpenAI for building multi-agent LLM workflows, including text agents, sandboxed workspace agents, and low-latency realtime voice agents. It provides primitives for agents, handoffs, tools (including MCP), guardrails, human-in-the-loop, automatic session history, and built-in tracing for debugging agent runs.

## Use cases
- build multi-agent workflows in typescript
- create realtime voice agents for spoken interactions
- orchestrate llm agents with handoffs and agents-as-tools
- add guardrails to validate agent inputs and outputs
- connect agents to tools via mcp or custom functions
- trace and debug agent runs
- manage conversation history across agent runs
- run agents in a sandboxed filesystem workspace

## When to choose
- You're building agentic AI applications in JavaScript/TypeScript on Node.js, Deno, or Bun
- You need multi-agent orchestration with handoffs, tools, guardrails, and sessions in one small package
- You want low-latency voice agents that connect to microphone and audio output in the browser via WebRTC
- You want built-in tracing and conversation history management without extra plumbing

## When to avoid
- You're building in Python and should use the Python OpenAI Agents SDK instead
- You need a heavyweight framework with extensive abstractions beyond a small set of primitives
- You need to run on edge runtimes like Cloudflare Workers as a first-class target (only experimental there)
- Your project doesn't involve LLM-driven agents or voice/text agent workflows

## Facets
- artifact type: framework
- maturity: active
- function: agent-framework, mcp, tracing
- domain: artificial-intelligence, large-language-models, speech-processing
- platform: bun, browser
- tags: typescript, openai, multi-agent-workflows, voice-agents, realtime, guardrails, handoffs, tool-calling, sessions, sandbox-agents, human-in-the-loop, webrtc, provider-agnostic, sdk, ai-agents, nodejs, deno

## Member repositories
- openai/openai-agents-js (main) score 81

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:08:15.183253+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T18:30:51.893189+00:00, confidence not recorded.
  - readme: https://github.com/openai/openai-agents-js (fetched 2026-08-28T04:08:15.183253+00:00, sha 4d57012337d9)
  - homepage: https://openai.github.io/openai-agents-js/ (fetched 2026-08-29T09:24:32.470232+00:00, sha 7528272ee935)
- Data as of 2026-08-30T08:39:29.467469+00:00.