openai/symphony
25,622 Stars·2,593 Forks·Elixir·Apache-2.0·11 Aufrufeopenai.com/index/open-source-codex-orchestration-symphony↗
Symphony
Symphony is an agentic workflow manager and autonomous software implementation engine. It serves as an orchestrator for large language model coding agents, converting high-level project requirements and task board items into verified pull requests.
The system manages an autonomous development workflow by delegating implementation runs to agents that handle end-to-end feature development and bug fixes. It generates automated pull requests backed by proof-of-work verification, ensuring that code contributions are validated before human review.
The platform coordinates a cycle of planning, coding, and verification using isolated run environments and state-based tracking to maintain progress. It integrates with task boards to translate project tickets into technical specifications and executable tasks.
Features
- Agentic Workflows - Serves as a coordination layer that manages the execution and verification of autonomous coding agents.
- Coding Agent Orchestrators - Coordinates multiple AI coding assistants to convert task board items into verified pull requests.
- Agentic Task Automation - Automates the transition from task board items to completed pull requests using autonomous agents.
- AI Software Engineering - Uses large language models to handle end-to-end feature implementation and bug fixes without constant supervision.
- Agentic Development Workflows - Provides a framework for managing project goals by delegating isolated runs to autonomous agents.
- Autonomous Agent Loops - Implements a continuous cycle of planning, executing, and verifying tasks to reach a defined goal state.
- Autonomous Software Engineering - Provides a platform for automating the full software development lifecycle from planning to verification.
- Isolated Execution Environments - Provides ephemeral sandboxes for implementation tasks to ensure reproducible results and prevent side effects.
- Autonomous Coding Agents - Orchestrates intelligent tools to iteratively write, test, and commit code until a feature is complete.
- Proof-of-Work Verifications - Validates completed tasks using automated tests and checksums to provide proof of work before PR submission.
- Hierarchical State Tracking - Maintains a hierarchical record of progress and attempted changes to allow agents to recover from failures.
- PR-Driven Delivery - Packages all autonomous implementation results into structured pull requests for final human review and integration.
- AI-Driven PR Generation - Implements project requirements and provides proof of work through automated code contributions in pull requests.
- Automated PR Generation - Turns project requirements into verified code submissions with integrated proof-of-work for review.
- AI Agents - Collaborative agent framework from OpenAI.
- Productivity and Collaboration - Tool for managing autonomous implementation runs and project work.
