output
The open-source TypeScript framework for building AI workflows and agents. Designed for Claude Code describe what you want, Claude builds it, with all the best
What is output?
The open-source TypeScript framework for building AI workflows and agents. Designed for Claude Code describe what you want, Claude builds it, with all the best practices already in place.
Overview of output
What it is
An open-source TypeScript framework for building AI workflows and agents. It is designed to be used with Claude Code, allowing users to describe what they want and have Claude build it with best practices already in place. The framework aims to consolidate prompts, evals, tracing, cost tracking, orchestration, and credentials into a single codebase, avoiding SaaS fragmentation and vendor lock-in.
Key features
- Build AI using AI: The codebase is structured so that AI coding agents like Claude Code can scaffold, plan, generate, test, and iterate on workflows. Each workflow is a folder containing code, prompts, tests, evals, and traces.
- Own your prompts: Uses
.promptfiles with YAML frontmatter and Liquid templating. Prompts are version-controlled, reviewable in PRs, and deployed with code. Providers can be switched by changing one line. - See everything that happens: Every LLM call, HTTP request, and step is traced automatically. Token counts, costs, latency, and full prompt/response pairs are stored as JSON in
logs/runs/. Zero config required. - Test AI like software: LLM-as-judge evaluators with confidence scores are available. Includes inline evaluators for production retry loops and offline evaluators for dataset testing.
- Use any model: Supports Anthropic, OpenAI, Azure, Vertex AI, and Bedrock through a single API. Structured outputs, streaming, and tool calling work consistently across providers.
- Scale without worrying: Uses Temporal under the hood for automatic retries with exponential backoff, workflow history, replay on failure, child workflows, and parallel execution with concurrency control.
- Keep secrets secret: Credentials are encrypted with AES-256-GCM, scoped per environment and workflow, and managed through the CLI. No external vault subscription needed.
Use cases
- Building AI-powered workflows such as blog post quality evaluation, sales call scoring, changelog generation, dependency auditing, recipe extraction, URL summarization, YouTube summarization, personalized Hacker News digests, and sales call processing.
- Creating production-ready AI agents that require orchestration, tracing, cost tracking, and multi-provider LLM support.
- Developing AI applications where complete ownership of prompts, data, and infrastructure is desired.
Getting started
Requirements: Node.js 20+, Docker Desktop, and an LLM API key (e.g., Anthropic).bashnpx @outputai/cli initcdnpx output devnpx output workflow run blog_evaluator paulgraham_hwhnpx output workflow debugAdd your API key to .env (e.g., ANTHROPIC_API_KEY=sk-ant-...). For the full getting started guide, see the repository documentation.
License and pricing
Licensed under Apache 2.0. The project is open-source and free to use.
Based on the project's public README and repository metadata.