OpenAI Symphony

Overview

OpenAI Symphony is an open-source orchestration tool for managing coding agents through ticket-based workflows and isolated workspaces. It is described as a background scheduler for agentic software delivery: instead of having an AI coding agent operate in an ad hoc way, Symphony continuously polls a work queue, such as Linear, and assigns structured tasks to agents based on predefined workflow rules.

For AI Product Managers, Symphony matters because it turns AI coding from a one-off assistant interaction into a repeatable delivery system. By combining ticket ingestion, workspace isolation, session lifecycle management, and standard operating procedures in `workflow.md`, it offers a practical operating model for scaling agent-driven software work while preserving process controls, review gates, and traceability.

Key Developments

• 2026-05-03: Jason Zhou shared a walkthrough showing how OpenAI Symphony, when paired with a codebase harness including Playwright CLI, Boot skill, and `WORKFLOW.md`, could improve coding agent outcomes by as much as 5×.
• 2026-05-04: A video demo showed how to set up and run OpenAI's open-source Symphony orchestrator with Linear tickets, using a `workflow.md` file and isolated workspaces. The coverage highlighted Symphony's background scheduler model, 30-second polling loop, per-ticket workspace creation, parallel agent limits, init hooks, and SOP-style markdown instructions for planning, validation, done criteria, and human-review triggers.

Relevance to AI PMs

1. Operationalizing agent workflows: Symphony gives AI PMs a concrete framework for moving from experimentation to production-like agent operations. Instead of manually prompting agents, teams can route work from ticket systems into governed execution flows.

2. Defining process and controls: The `workflow.md` pattern is especially useful for PMs because it encodes scope, filters, validation steps, and escalation rules in a durable artifact. This makes it easier to specify how agents should behave, when humans should review output, and what “done” means.

3. Improving throughput without losing oversight: By using isolated workspaces and configurable parallelism, Symphony supports multiple active agent sessions while reducing interference across tasks. For PMs managing delivery velocity, this provides a way to experiment with agent scaling while keeping auditability and task boundaries intact.

Related

• Linear: A key integration point for Symphony's ticket-based orchestration. Symphony polls Linear projects and selects eligible “to-do” items for agent execution.
• workflowmd: Closely connected to Symphony's operating model. The `workflow.md` file defines project configuration and SOP-style instructions that guide agent behavior.
• playwright-cli: Often used as part of a codebase harness alongside Symphony to validate UI behavior and strengthen execution reliability.
• boot-skill: Referenced as another component in the supporting harness that can improve coding agent outcomes when combined with Symphony.
• jason-zhou: A prominent early source demonstrating Symphony setups and explaining its practical workflow.
• claude-code: Related as another coding-agent tool in the broader ecosystem, useful for comparison in agentic software delivery workflows.
• openai: The organization associated with Symphony as an open-source orchestrator.
• playwright-cri: Related through the broader Playwright-based tooling ecosystem used for testing and automation around coding agents.