refly-ai/refly

Project overview

Refly converts natural-language business logic into structured, version-controlled agent skills that can be invoked reliably. It provides an intervenable runtime to pause and re-steer agents, along with a skill registry for secure management and sharing.

Project type

AI Agent · Workflow

Use cases

Automation

Deployment

Refer to project documentation

License

License pending

Best for

• Developers and enterprise teams who need to standardize and version agent skills for production automation.

Key capabilities

• Allows users to pause, audit, and re-steer agent logic mid-run.
• Securely manage, version, and share agent capabilities.

Limitations and risks

• Configuring a chat model from an external provider, such as OpenAI or Anthropic, is required.
• GPU, minimum hardware, and database requirements are not documented.
• Setup difficulty is rated medium due to the need for self-deployment and model provider configuration.

Getting started

• Deploy Refly using Docker, navigate to localhost:5700, register an account, configure an external LLM provider, and create a workflow.

Evidence and sources

• GitHub project description: The first open-source agent skills builder. Define skills by vibe workflow, run on Claude Code, Cursor, Codex & more.
• README: Skills are not prompts. They are durable infrastructure.
• README: Most AI Agents fail in production because they rely on "Vibe-coded" scripts and fragile, black-box logic.
• README: Step-by-step guide to deploying Refly on your own server using Docker.
• README: Complete API documentation for integrating Refly into your applications.

Reviewed at: 2026-07-16T23:34:49Z