Sylph is an open-source platform for building a company brain that runs business operations through AI agents, skills, and self-improving context.
Sylph addresses the challenge of coordinating AI agents across an entire organization by providing a structured repository system where company knowledge, processes, and agent definitions live together. The approach centers on storing domain-specific context in organized folders, defining reusable skills as self-improving instruction files, and configuring AI agents as virtual employees with assigned capabilities. The core mechanism is a self-learning loop: after you approve an agent's output, the system diffs what was generated against what you kept, then rewrites the skill's rules to better match your preferences in the next cycle. The tool works agent-agnostic, integrating with Claude Code, Codex, Cursor, or any AI coding agent that can operate on a git repository.
Sylph suits teams wanting to automate cross-functional workflows without building custom infrastructure. It works across all business domains—product, marketing, sales, legal, recruiting—and handles tasks ranging from content creation and event planning to report generation and task prioritization. The platform is designed for collaborative team use, allowing multiple people to contribute to shared knowledge bases. You should adopt it if you want agents that improve their own performance over time through your feedback, rather than requiring constant manual prompt refinement. The tool is particularly suited to startups and small organizations where a single repository can serve as the operational backbone.
Development on the project shows active engagement with the codebase through regular commits and ongoing refinement of core features. The maintainers continue to expand skill capabilities and agent configurations based on real operational use. Documentation is detailed and includes practical examples of how to set up and customize agents for specific company needs.
