OpenAI's Symphony orchestrates issue-tracker coding agents
Symphony is OpenAI’s open-source spec and reference implementation for orchestrating coding agents around project-tracker tasks instead of individual chat sessions. The system watches a board like Linear, creates isolated per-issue workspaces, runs agent sessions continuously, and hands back proof of work such as CI status, review feedback, complexity analysis, and walkthroughs. The repo is language-agnostic at the spec level, ships an experimental Elixir implementation, and is explicitly framed as a trusted-environment preview rather than a hardened general-purpose platform.
Hot take: this is not “local LLM support” out of the box; it is an operating model for agentic software delivery that you could adapt to local models if your harness already speaks the right runtime and tools.
- –The core idea is orchestration, not model quality: Symphony treats the issue tracker as the control plane and the agent as an execution worker.
- –The repo is intentionally spec-first and language-agnostic, which makes it easier to reimplement against a local harness than to “configure” like a normal app.
- –The published reference implementation is Elixir-based, so anyone pairing it with OpenCode, pi, or another local stack should expect integration work around sessions, workspaces, and tool execution.
- –OpenAI is positioning it for trusted environments with clear guardrails, so this reads more like a systems pattern than a turnkey enterprise product.
DISCOVERED
4h ago
2026-04-29
PUBLISHED
7h ago
2026-04-28
RELEVANCE
AUTHOR
valtor2