openai/symphony

Symphony is a tool that lets software teams delegate coding tasks to AI agents instead of micromanaging them step by step. The core idea is to shift engineers from "supervising" a coding agent to simply "managing work", you describe what needs to be done, Symphony spins up isolated agent runs to handle each task, and the agents take care of the actual implementation. Here is how it works in practice: Symphony can monitor a project management board (the demo shows it watching a task board called Linear), automatically pick up work items, and spawn separate AI coding agents to tackle each one independently. The agents don't just write code, they provide "proof of work" when they finish: things like automated test (CI) status, code review feedback, complexity analysis, and even walkthrough recordings. Once a task is approved, Symphony handles merging the code safely. You would reach for Symphony if your engineering team already uses AI coding tools and wants to stop babysitting each individual session. Instead of staying in the loop for every decision, you set the work queue and review the results. It is described as an "engineering preview," meaning it is early-stage and suited to trusted internal environments rather than public production use. The reference implementation is written in Elixir, but the project also provides a specification document so you can build your own version in any language using your preferred coding agent. It is released under the Apache 2.0 open-source license.