Symphony turns project work into isolated, autonomous implementation runs so teams can manage work instead of supervising individual coding sessions.
In this demo video, Symphony monitors a tracker board for work and starts isolated implementation runs for selected tasks. Each run produces proof of work such as CI status, PR review feedback, complexity analysis, and walkthrough videos. When accepted, Symphony lands the PR safely. Engineers do not need to supervise individual coding sessions; they can manage the work at a higher level.
Warning
Symphony is a low-key engineering preview for testing in trusted environments.
- Claude support: Agents can run on claude as well as codex.
- Github issues: In addition to Linear, agents can watch and move Github issues.
- Tracker adapters: configure tracker backends for board- or issue-based queues, including label-based state machines where the tracker supports them.
- Implementation adapters: configure implementation backends through Symphony's app-server protocol.
- Live run logs: each workspace writes
logs/agent.mdandlogs/agent.ndjson; the dashboard can open those logs in a live-updating modal while a run is active. - Dashboard auth and hosting: the Phoenix dashboard supports Basic Auth and can be bound to a configured host/port for private operational access.
- Workflow helpers: repo-local skills and scripts keep issue work, PR creation, and landing behavior consistent across runs without making those workflows part of Symphony's core model.
See elixir/README.md for the supported WORKFLOW.md options and
adapter examples.
Symphony works best in codebases with clear setup instructions, automated validation, and workflow conventions that autonomous implementation runs can follow.
See elixir/README.md for setup, configuration, and the agents command
reference (foreground, background, and stop modes on Linux and macOS).
This project is licensed under the Apache License 2.0.