AgentBrew is an open-source project that provides a portable toolbelt for AI agents, abstracting MCP servers and tools to prevent agent and tool lock-in across different frameworks.
Hi everyone, I’m working on an open-source project called AgentBrew to solve a frustration I think many of us building with AI agents are running into right now: Agent & Tool Lock-In. **The Problem** The agent ecosystem is exploding. One day you're building with Claude Code or Goose, the next day you're experimenting with a custom LangGraph or AutoGen setup. Every time you switch frameworks or spin up a new agent, you have to reinvent the wheel configuring your Model Context Protocol (MCP) servers, local tools, and agent skills. Your tools are tightly coupled to the specific agent framework you wrote them for. **The Solution: AgentBrew** AgentBrew acts as a portable toolbelt for your AI agents. Instead of binding your MCP servers, custom tools, and skills directly to a single agent configuration, AgentBrew abstracts them. **Share Your Thoughts!** This project is under active development, and I would love to get the community's perspective on it. Please check out the repo, try it out in your workflow, and share your experience with me! I'm incredibly eager to hear your honest feedback, what you feel is currently lacking, and any suggestions you have to make the tool better. Feel free to open an issue on GitHub or drop a comment below.
AgentBuddy is a local-first, open-source AI workflow sandbox that enables persistent agent threads, real-time execution traces, and event-driven workflows, with Claude Code integration, aiming to keep AI development local and transparent.