I built Agent‑Flow, an MCP server for managing and using reusable prompts and interactive workflows with AI agents.
- Prompts: reusable instructions that save you from repeating the same request.
- Workflows: guided, multi‑step processes with branches and loops. For users, they eliminate repeated manual inputs. For agents, they enable complex tasks by dynamically loading the right prompts into the context window.
Agent‑Flow works with any MCP host — such as Copilot, Claude‑Code, Cursor, or Gemini CLI — and runs in the public cloud. The web console makes onboarding simple: you create a profile, configure settings, and manage collections of prompts and workflows. Collections can also be maintained via GitHub, and public ones are listed in the community repository for discovery through the console.
sileo-oss•1mo ago
I built Agent‑Flow, an MCP server for managing and using reusable prompts and interactive workflows with AI agents.
- Prompts: reusable instructions that save you from repeating the same request.
- Workflows: guided, multi‑step processes with branches and loops. For users, they eliminate repeated manual inputs. For agents, they enable complex tasks by dynamically loading the right prompts into the context window.
Agent‑Flow works with any MCP host — such as Copilot, Claude‑Code, Cursor, or Gemini CLI — and runs in the public cloud. The web console makes onboarding simple: you create a profile, configure settings, and manage collections of prompts and workflows. Collections can also be maintained via GitHub, and public ones are listed in the community repository for discovery through the console.
GitHub community repo: https://github.com/sileo-oss/agentflow-community
I’d love to hear your feedback — what do you think about reusable workflows for AI agents, and what features would make Agent‑Flow more useful to you?