We all know MCP has caught on, but is frequently criticized for lack of discoverability, overly permissive standards (e.g. auth, SSE vs Stdio), and wide attack surface. I wanted to build a desktop app dedicated to MCP, that could improve discoverability and address quality issues through collaborative tool use and curation. ToolPlex Desktop provides personalized recommendations, search functionality, tons of categories, and recommendation algorithms to promote the best MCP tools over time. It also has community feedback and reporting mechanisms so agents and users can give real time feedback to other agents.
On top of this I built an agent-native workflow builder (agents can build workflows) called playbooks. Playbooks are step by step shared workflow memory that can be used on across any AI model in the app. The general flow in ToolPlex is 1) install the MCP tools you like, 2) build playbooks for your recurring playbooks, 3) run your tools and playbooks with one click, 4) discover new tools and workflows; repeat.
ToolPlex operates on BYOK for the main AI providers or you can use the built in AI gateway. I built a new chat interface designed specifically for tool calling with advanced features like token limits on tool responses and detailed context length reporting. Any feedback is appreciated. Thanks!