OpenAI Releases Symphony: Autonomous Agents That Convert Project Tasks Into Working Code

Summary

OpenAI has released Symphony, an experimental framework that autonomously converts project management tasks into completed code implementations. The system monitors project boards like Linear or Jira, spawns AI agents to handle individual tasks, and completes the full development cycle—from writing code to creating pull requests, running CI tests, and providing walkthrough videos. Unlike traditional coding assistants that require constant human supervision, Symphony operates at a higher abstraction level, allowing engineering teams to manage work outcomes rather than supervise individual coding steps.

The framework represents a shift from "managing coding agents" to "managing work that needs to get done," according to OpenAI's documentation. When a task appears on a monitored board, Symphony automatically creates isolated implementation runs where agents write code, submit PRs with proof of work including CI status and complexity analysis, and safely merge changes once approved. The system is designed for codebases that have adopted what OpenAI calls "harness engineering"—infrastructure that supports autonomous agent operation.

Released as an open-source project under Apache License 2.0, Symphony is currently labeled as a "low-key engineering preview" intended for testing in trusted environments. OpenAI provides both a specification document and an experimental Elixir-based reference implementation, though the company encourages developers to build their own versions in any programming language using the provided spec. The release signals OpenAI's vision for AI agents that can handle entire feature implementations with minimal human intervention, fundamentally changing how software development work is organized and executed.

• Requires codebases with "harness engineering" infrastructure and is currently an early preview for trusted environments only
• Represents a paradigm shift from supervising AI coding tools to managing work outcomes at a higher level of abstraction

Editorial Opinion

Symphony represents an ambitious leap toward truly autonomous software development, but its success will hinge on real-world reliability and error handling. While the vision of converting Jira tickets directly into merged code is compelling, the "trusted environments only" caveat suggests OpenAI knows this technology isn't yet ready for production use at scale. The requirement for "harness engineering" infrastructure also means most teams will need significant upfront investment before they can even experiment with Symphony, potentially limiting early adoption to well-resourced organizations already invested in AI-native development practices.